Data augmentation

Overview

Labelled data is expensive, and a model often sees too few examples of some classes or situations. Data augmentation adds training signal without new labelling. Most variants transform existing examples. The invariant they all need is that the transformed example still has a correct label, either the original one or one that has been updated to match the transform 2012.07177 §3 1802.09567 §4.2. One variant works on embeddings rather than pixels, and there the invariant is that sampled vectors stay near the class's estimated distribution 2101.06395 §3.2.2.

Several variants depart from the transform pattern. Augmented SBERT builds new sentence pairs and takes their labels from a teacher model, the generative variant makes new images from class-name prompts, and the model-filtered sentence variant keeps candidate sentences that a tagger agrees with 2606.29836 §3.3. Two more generate or insert data: a per-type GAN makes synthetic failure rows for tabular sensor data, and a forecaster inserts a virtual value between each pair of real readings 2607.19153 §3.2.5 2512.06714 §3.5. A survey also describes language models writing training examples 2111.01243 §5. Two further papers generate whole training items: a multimodal model writes navigation instructions for routes built from house-tour videos, and a diffusion model makes LiDAR objects that are then placed in scenes 2411.11394 §3.2 2412.17226 §3.2. They are included because their papers present them as augmentation or training-data generation, and their sections say where the invariant differs.

The contrastive variants (graph, time-series and tabular pre-training) use no label in their pre-training step. Their invariant is that two views of one sample should agree 2106.14112 §3.3 2103.00111 §6 2106.01342 §4.

The variants on this page differ in what they change:

• whole images composited from parts of two images, with object masks rewritten to match 2012.07177 §3;
• a patch cut from one image and pasted onto another at a saliency peak, with the label weights set by the patch area 2006.01791 §3.1 2006.01791 §3.2;
• individual images or characters, with flips, colour inversion or a fixed set of geometric and photometric transforms 1802.09567 §4.2 1811.04241 §4 2009.08369 §4.1. A survey of 40 agricultural studies counts rotations in 12 papers and gives no parameter ranges 1807.11809 §4.5;
• patches cut from large microscope images, rotated by 90, 180 and 270 degrees, with the sampling stride set per class 1706.06480 §4.3.4;
• feature vectors sampled from a class Gaussian that is borrowed from similar base classes 2101.06395 §3.2.2;
• repeated copies of scarce satellite images, each with its own random transform 2201.09206 §3.3;
• graph structure and node features, used to make the views that contrastive pre-training compares 2103.00111 §6.1;
• a weakly and a strongly perturbed copy of each raw time series, used as the two views of a contrastive pretraining step 2106.14112 §3.1;
• a bundle of strong image-level policies used when training vision transformers from scratch 2202.06709 app. A;
• sentence pairs recombined from the labelled set and scored by a cross-encoder 2010.08240 §3.1;
• images drawn from a fine-tuned text-to-image model, one prompt per class 2304.08466 §4.1;
• tokens in a tagged sentence, replaced or shuffled so that the BIO tags stay valid 2010.11683 §3;
• sentences from unlabelled abstracts, tagged by a model and kept only when the model finds every known entity in them 2606.29836 §3.3;
• patch tokens of a vision transformer, each with a dense label from an offline annotator, mixed as patch embeddings 2104.10858 §3.3;
• pixels of a segmentation model, mixed in the loss only, with unblended label maps 2107.12283 §6;
• offline rotated and noisy copies of a small photo set 2605.11989 §3.3.3;
• perturbed copies of synthetic temporal graphs, with facts dropped, relations renamed, entities renamed and years shifted 2401.06853 §3.2.2;
• mosaic and mixup as named steps of detector training, with no parameters given 2408.15857 §3.1.1 2408.15857 §3.2;
• label-safe transforms drawn on the fly for whole mammograms: flips, rotations, shear, zoom and channel shifts 1708.09427 §3.1.2;
• reflections, shifts, scales and shears that expand a small phone-photo set nine-fold before the split 2606.23286 §5.1;
• flips and rotations that grow a grain photo set sixfold 2609.10524 §4.1;
• synthetic failure rows from one WGAN-GP per failure type, pooled with the real minority rows 2607.19153 §3.2.5;
• one virtual value inserted between each pair of consecutive demand readings 2512.06714 §3.5;
• training examples that a language model writes, as a survey describes them 2111.01243 §5;
• a heavy chain of rotation, zoom, shear, colour and brightness changes, and flips, for an ImageNet-pretrained classifier on field photos 1810.05726 § Deep Learning;
• CutMix on table cells and mixup on the row embeddings, as views for contrastive pre-training 2106.01342 §4;
• resampling of minority rows, by SMOTE and its variants, ADASYN, random over- or under-sampling, or a GAN, before a classifier is fitted 2304.02858 §4.1 2304.02858 tab. 1.
• virtual raindrops ray-traced onto real clean street photos, with the drop mask taken from the simulation 2609.37870 §3.
• a batch-level transform pipeline in a deep learning library, run on the GPU, with rotation, zoom and lighting defaults 2002.04688 §2.1;
• node and edge drops for graph contrastive learning, with probabilities set by node centrality so that influential structure survives 2010.14945 §3.3.1 2010.14945 §3.3.2;
• RandAugment and label smoothing inside a long closed-set training recipe, used for open-set recognition 2110.06207 app. D;
• cached Mosaic and MixUp in detector training, followed by a final stage with large-scale jittering 2212.07784 §3.3 2212.07784 tab. 7;
• a second frame made by warping an annotated still image along a smooth flow field, so that video-style training pairs exist without filming 1702.00871 §4;
• edge drops whose probabilities come from a learned network, trained to make two views of a graph differ, with a penalty on how many edges go 2106.05819 §3.2;
• robot demonstrations re-rendered with another arm and gripper in the same pose, with the background filled in behind the original arm 2512.13100 §4.1 2512.13100 §4.2;
• route-and-instruction pairs, one per house-tour route, with the instruction written by a multimodal model and checked against the route's room and action labels 2411.11394 §3.2;
• LiDAR foreground objects from a text-and-box diffusion model, placed in a range-image scene model with a per-category loss 2412.17226 §3.2 2412.17226 §3.3;
• random 3D rotation, scaling and shear of skeleton joint coordinates, applied to each training sequence 1704.02581 §4.3;
• a random gamma-type brightness curve on synthetic indoor images, with the depth map left unchanged 2408.06083 §3.1;
• real face crops stitched onto other identities inside elastically deformed masks, which are known because the method makes them 2012.09311 §3.2.

Creating new labelled scenes from 3D meshes is a related but different method. It is covered under scope below. Test-time augmentation averages predictions over transformed inputs and changes no training data. It is not a training method, so it is covered under scope below too.

Variants

Copy-Paste compositing for instance segmentation

Two random training images are each scale-jittered and flipped. A random subset of the objects from one image is pasted onto the other, and the objects of the receiving image that are covered are trimmed or removed from the labels 2012.07177 §3. Paste locations are random and no scene context is modelled 2012.07177 §3. Card: Ghiasi et al. (Copy-Paste).

Reported gains on COCO for Mask R-CNN-style models are +1.0 to +1.5 box AP and +0.5 to +1.1 mask AP across seven models 2012.07177 tab. 1. With self-training on top, the strongest model reaches 57.3 box AP and 49.1 mask AP on test-dev 2012.07177 tab. 4. The gain is largest when labels are scarce: on 10% of COCO it adds 6.9 box AP over scale jittering alone and 4.8 over large scale jittering 2012.07177 §4.3.

The paper's ablation finds that pasting a random subset beats pasting a single object or all objects 2012.07177 tab. 10. Blending the paste edges made no difference to AP 2012.07177 tab. 10. Random pairing often produces scenes that do not match, such as indoor and outdoor images mixed in about 47% of generated examples 2012.07177 app. C 2012.07177 tab. 12.

Saliency-guided patch mixing for classification

A patch is cut from a source image and pasted at the same location on a target image. The patch is centred on the most salient pixel of the source, found with an OpenCV saliency detector, and the two labels are mixed in proportion to the patch area 2006.01791 §3.1 2006.01791 §3.2 2006.01791 §3.3. Among five placement schemes, pasting at the corresponding location gives the lowest error in the paper's experiments 2006.01791 §3.4. The motivation is that a random patch often lands on background, and the mixed label would then credit background to the source class 2006.01791 §1. Card: Uddin et al. (SaliencyMix).

For WideResNet-28-10 without traditional augmentation, top-1 error falls from 5.18% to 4.04% on CIFAR-10 and from 23.21% to 19.45% on CIFAR-100, against CutMix 2006.01791 tab. 1. With crop and flip added, the gap is much smaller: 2.87% against 2.76% on CIFAR-10 and 16.66% against 16.56% on CIFAR-100 2006.01791 tab. 1. On ImageNet, ResNet-50 moves from 21.40% to 21.26% top-1 over CutMix 2006.01791 tab. 2. The ImageNet numbers are the best run per method, while the CIFAR numbers are means over five runs, so they are not directly comparable 2006.01791 §4.1.2 2006.01791 tab. 1.

The released code differs from the paper in ways that change the augmentation. It computes one saliency box per batch, from the first partner image, and pastes that box from every partner saliencymix.py. Its box side is the square root of 1 − λ times the image side, and it recomputes the label weight from the clipped area saliencymix.py saliencymix.py. The ImageNet script trains on a 200-class Tiny-ImageNet path train.py, so the ImageNet-1k numbers need more than the released script.

Rotated patches with class-balanced sampling for pixel-wise segmentation

For pixel-wise segmentation of large microscope images, the training inputs are 1000 × 1000 px patches cut from the full image with a sliding window 1706.06480 §4.2 1706.06480 §4.3.3. Each patch is rotated by 90, 180 and 270 degrees 1706.06480 §4.3.4. The window stride is set per class so that each class yields about the same number of patches, which gives rarer classes a smaller stride 1706.06480 §4.3.4. The exact stride values are not stated. The worked example uses a 100 px stride on a 7000 × 8000 px image, which gives 4331 patches 1706.06480 §4.3.4. Card: Azimi et al. (steel microstructure).

The paper's ablation adds the two steps one at a time. On SEM images with a fine-tuned fully convolutional network, balancing alone moves pixel accuracy from 92.01% to 92.11%, and rotation on top of balancing moves it to 93.92% 1706.06480 tab. 5. Per-class pixel accuracy does not move in one direction: balancing alone lowers tempered martensite from 71.43% to 57.44%, and rotation brings it back to 72.62% 1706.06480 tab. 6. At the object level, the full method reaches 93.94% from 90.97% with balancing alone 1706.06480 tab. 2. The text gives the augmentation gain as 2 points 1706.06480 §6; the table difference is 2.97 points.

The rotation is a label-preserving assumption. The paper reports similar responses to rotated patches and calls the system rotation invariant 1706.06480 §6, but it does not test invariance of the architecture. All numbers come from one image-level split of 21 images, with no repeated runs 1706.06480 §5.1. Use this only where a 90-degree turn leaves the label unchanged.

Feature-space sampling for few-shot classifiers

This variant does not change images at all. The feature extractor is frozen. For each support feature, the method borrows a Gaussian from the two nearest base classes, averages it with the support feature's own statistics, and draws 750 vectors per class from it 2101.06395 §3.2.2 2101.06395 §3.2.3. A logistic regression is trained on the support features plus the drawn vectors 2101.06395 §3.2.3. Card: Yang et al. (distribution calibration).

Before sampling, the features are passed through a power transform with exponent 0.5 2101.06395 §3.2.2. The method assumes non-negative features, such as pooled post-ReLU features 2101.06395 §4.1.3.

In the authors' miniImageNet 5-way ablation, the two steps add to each other for 1-shot accuracy 2101.06395 tab. 4:

The CUB value of α in the paper is not the one the released code uses, and the paper's tieredImageNet results are not covered by that code 2101.06395 §4.1.3 2101.06395 §4.1.1. Treat the cross-dataset numbers as the authors' results only.

Repeated augmented copies of scarce classes

The University-1652 benchmark has one satellite image per building, so training sees a few satellite images for each class. The fix is to draw each satellite image k times in a training step. Every copy gets its own random transform, such as shifting, padding, cutting or colour change 2201.09206 §3.3. Card: Dai et al. (FSRA).

With the triplet loss alone, drone-to-satellite AP rises as the copy count goes up 2201.09206 tab. 7:

The extra copies make training slower but leave inference unchanged 2201.09206 §3.3 2201.09206 §4.4. The best k depends on the batch size: the paper reports k = 3 at batch 8 and k about 5 at batch 16, and the batch-16 values appear only in a figure 2201.09206 app. B.

Graph augmentations for contrastive self-supervision

For graphs, a view is a perturbed copy of a node, subgraph or whole graph. Attribute-level views mask or shuffle node features. Topology-level views modify edges, apply graph diffusion, or sample a subgraph 2103.00111 §6.1. The model is trained so that two views of the same object agree and views of different objects do not 2103.00111 §6. Card: Liu et al. (graph SSL survey).

The survey reports that contrastive methods using augmented views (MVGRL, MERIT, SubG-Con) outperform generation-based and auxiliary-property methods on node classification 2103.00111 §8 2103.00111 tab. 5. It also notes that augmentation choices in contrast-based methods are tuned empirically and are costly to tune 2103.00111 §3.2. The augmentation strengths used by the surveyed methods are not stated 2103.00111 §8.

Centrality-adaptive drops for graph contrastive learning

Two views of a graph are built by removing edges and masking feature dimensions. Each edge and each feature dimension gets a removal probability from a score: edges from the centrality of their endpoints, feature dimensions from the centrality-weighted sum of the nodes that have them 2010.14945 §3.3.1 2010.14945 §3.3.2. Low-scoring items are removed more often, and a cap keeps any item from being almost always removed 2010.14945 eq. 4. Degree, eigenvector and PageRank centrality are the three measures tried 2010.14945 §3.3.1. Card: Zhu et al. (GCA).

The uniform version of the same two steps is the baseline. On five node classification datasets the adaptive scheme beats uniform dropping in every case, by 0.1 to 1.6 accuracy points, and the largest gap is on Amazon-Computers 2010.14945 tab. 4. The paper's own reading is that each adaptive part helps and that both together help most 2010.14945 §4.3. The rules and the cap are described in the paper, but the official code differs from the paper in edge weighting and in the PageRank update, so reimplementations should check 2010.14945 §3.3.1. Those differences are recorded on the card.

Learned edge drops for graph contrastive learning

Here the drop probabilities are not set by a rule. A graph neural network (the augmenter) outputs one drop logit per edge. The encoder is trained to match each graph with its edge-weighted copy, and the augmenter is trained to make those copies differ. A penalty on the mean drop probability per graph keeps the augmenter from dropping nearly everything 2106.05819 §3.2 2106.05819 eq. 8. Card: Suresh et al. (AD-GCL).

The paper argues that the min-max form links the objective to a graph information bottleneck, so the encoder keeps less information that the task does not need 2106.05819 §3.1. The theorem behind this assumes a 1-WL-powerful encoder and a countable graph space, and the paper gives no guarantee for a trained network 2106.05819 §3.1 2106.05819 app. B. Gradients through the discrete edge choice use a Gumbel-max relaxation, so the drop is soft during training 2106.05819 §3.2.

The authors' unsupervised results are linear-probe accuracies over 10 seeds. On the TU collection the paper counts significant gains over InfoGraph and GraphCL on 5 of 8 and on 6 of 8 datasets 2106.05819 §5.1. On RDT-M5K the value is 54.93 for AD-GCL-OPT against 53.05 for GraphCL 2106.05819 tab. 1. The fixed-λ and tuned-λ versions of the method give similar numbers, which the paper reads as robustness to the penalty weight 2106.05819 §5.1. On transfer to chemistry and biology tasks, AD-GCL-FIX has a mean rank of 2.4 across nine datasets, with the baseline numbers copied from earlier papers 2106.05819 §5.3 2106.05819 app. G.

The regulariser sets the drop ratio. For classification, λ from 0.1 to 10 gives expected drop ratios from about 80% down to under 10%, and accuracy is flat across that range 2106.05819 §5.2 2106.05819 fig. 3. For regression, the paper finds a smaller penalty, and so more dropping, works better 2106.05819 app. F. Edge dropping is a proof-of-concept search space; the paper says other families may do better 2106.05819 §3.2.

The released code differs from the paper in several details, among them how dropped edges are handled, the contrastive temperature and the default dropout 2106.05819 app. G. Those differences change the augmentation, so check the card's Paper vs code section before reproducing the numbers.

Weak and strong views for time-series contrastive pretraining

Time series have no labels in the pretraining stage, so the augmentation only decides which two views the model must match. TS-TCC makes a weak view with jitter and scaling. It makes a strong view by cutting the series into a random number of segments, shuffling them, and adding jitter. The paper says the segment count should grow with sequence length, and that jitter should be smaller for normalised data 2106.14112 §3.1. The two views feed a temporal contrasting module and a contextual contrasting loss 2106.14112 §3. Card: Eldele et al. (TS-TCC).

The ablation on the three labelled tasks shows the two views doing different work. On HAR, the weak view alone gives 76.55% accuracy and the strong view alone gives 60.23%, against 90.37% for both 2106.14112 tab. 4. On Sleep-EDF the single views give 80.90% and 78.55%, against 83.00%. On Epilepsy the single views give 97.18% and 97.14%, against 97.23%. The text reads this as one view alone hurting HAR and Sleep-EDF, but not Epilepsy 2106.14112 §5.4. All figures are author-reported over five seeds 2106.14112 §4.2.

The linear probe on the frozen encoder reaches 90.37% on HAR, against 90.14% for the supervised model. On Sleep-EDF it reaches 83.00% against 83.41% supervised, and on Epilepsy 97.23% against 96.66% 2106.14112 tab. 2. Strong jitter strength is not stated in the paper, and the released code uses different segment counts and scales from the text (see the card's Paper vs code). The released code also samples the augmented views once per run, so each sample keeps the same two views in every epoch dataloader.py. Re-sampling per epoch would be a different method. Read from the code, the strong permutation may give every channel the same ordering on multichannel data augmentations.py. Check this before reusing the strong view on data such as HAR's nine channels.

Strong views matched through similarity distributions

Keep the weak view as a teacher, and ask the strong view to reproduce the weak view's similarity profile over a queue of stored keys, rather than to match its own key 2104.07713 §3.2 eq. 6. The term is added to a standard contrastive loss with weight β = 1.0, with τ = 0.2 and key momentum 0.999 2104.07713 eq. 7 2104.07713 §3.3. The strong view is built from 14 operations, drawn in five rounds with probability 0.5 each 2104.07713 §3.3 tab. 1. Card: Wang & Qi (CLSA). Author-reported ImageNet linear top-1 at 200 epochs is 69.4% against 67.5% for MoCo v2 2104.07713 tab. 2. Without the distribution term, a strong query alone gives 67.7 and a strong query with a strong key gives 67.0 2104.07713 tab. 5. The wall-clock cost is 70 h against 53 h for MoCo v2 at 200 epochs 2104.07713 tab. 6.

Strong augmentation in vision-transformer training

Training a vision transformer or a hybrid from scratch on small data often uses a bundle of strong image-level policies. The paper names RandAugment, Random Erasing, label smoothing, mixup, CutMix and stochastic depth, with AdamW and a cosine schedule 2202.06709 app. A. Card: Park & Kim (ViT analysis). Most values for these components are not stated in the paper, so a reimplementation has to choose them 2202.06709 app. A.

The paper's own analysis gives mixed effects. Strong augmentation reduces the size of the Hessian eigenvalues but produces many negative ones 2202.06709 app. F. Models trained without it are over-confident on CIFAR-100, and models trained with it become under-confident 2202.06709 app. F. The card records no accuracy ablation of this bundle on its own 2202.06709 app. A.

RandAugment and label smoothing in a long closed-set recipe

For open-set recognition, the paper trains a closed-set classifier with a strong recipe and then scores unknowns with the maximum logit. The recipe includes RandAugment, with its count and magnitude tuned per dataset, and label smoothing on TinyImageNet only, at 0.9 2110.06207 app. D. In the official code the smoothing value is the mass taken away from the true class, so the true class keeps only 0.1 of the target 2110.06207 app. D. On TinyImageNet, switching from crop and flip to RandAugment at 200 epochs adds 1.8 AUROC points, and adding label smoothing at 600 epochs adds 4.6 more 2110.06207 app. C. These are single-setting ablations on five known/unknown splits. Card: Vaze et al. (open-set recognition).

Recombined sentence pairs with teacher labels

Sentence-pair models can learn from pairs that nobody labelled by hand. A slow, accurate cross-encoder is trained on the gold pairs. It then scores new pairs made by recombining gold sentences, and a fast bi-encoder is trained on the gold pairs plus these soft-labelled pairs 2010.08240 §3.1. The cross-encoder's score is the label, so the bi-encoder learns from soft targets 2010.08240 §3.1 2010.08240 §7. Card: Thakur et al. (Augmented SBERT).

Which pairs are sampled decides the gain. The paper compares random sampling, kernel density estimation (KDE) that matches the gold score distribution, BM25 lexical neighbours, semantic-search neighbours, and a merge of the last two 2010.08240 §3.1. On the five in-domain datasets, the best strategy beats the lower-bound SBERT baseline by about 1.0 to 6.0 points. Random sampling falls below that baseline on Spanish-STS, at 62.05 against 72.07 2010.08240 tab. 5. BM25 gives the smallest silver set in every row, for example 3,964 pairs on Spanish-STS against 230,364 for KDE 2010.08240 tab. 11.

For domain adaptation, no labelled target data is needed. The cross-encoder is trained on a labelled source domain and labels the unlabelled target pairs 2010.08240 §3.2. The largest reported change is Quora to Sprint, where AUC(0.05) rises from 0.505 to 0.875 2010.08240 tab. 6. In most other rows the result stays below the in-domain upper bound 2010.08240 tab. 6.

This is closer to pseudo-labelling than to transforming an example. The new pairs are not transformed copies of gold pairs, and their validity rests on the teacher's scores being useful labels. It is listed here because the paper presents itself as data augmentation.

Generated images from a fine-tuned text-to-image model

Instead of transforming existing images, this variant makes new labelled images with a generator. The authors fine-tune Imagen on the ImageNet-1K training images, prompt it with each class name, and generate as many images per class as there are real ones, at scale factors from 1 to 10 2304.08466 §4.1 2304.08466 §4.3. The label is the class in the prompt, so label validity depends on the generator producing that class. Card: Azizi et al. (synthetic ImageNet).

Classifiers trained only on generated images do worse than real-only training in every row of the classifier table. A ResNet-50 reaches 64.96% top-1 on real validation images with generated data at 256 × 256, and 69.24% at 1024 × 1024, against 73.09% for real data under the CAS recipe 2304.08466 tab. 2. Adding 1.2M generated images to the real training set raises ResNet-50 top-1 from 76.39% to 78.17% 2304.08466 tab. 3.

The gain does not grow with the synthetic set. In the scaling table, the best total size is 2.4M images at both resolutions, with +1.22 at 256 and +1.73 at 1024. At 6.0M total the 256 result is 0.30 below the real-only baseline, and the 1024 result is 0.05 below 2304.08466 tab. 4. The table's size axis is read as the total training set, since its 1.2M row equals the real-only baseline; the paper does not define the column 2304.08466 tab. 4.

Fine-tuning used 256 TPU-v4 chips for 210K base steps and 490K steps for the 64-to-256 super-resolution stage, at batch 2048 2304.08466 §4.1. The generation cost of the 1.2M to 12M sets is not reported 2304.08466 §4.3. The paper links no code, and its generator is not linked from the paper either 2304.08466 §4.1.

Per-type GAN oversampling for tabular failure data

This variant synthesises minority rows for tabular sensor data. Instead of one generator for all failures, it trains one per failure type. Failure rows are split by type (HDF, PWF, OSF, TWF), a WGAN-GP is fitted to each group with its own Optuna study, and each generator draws samples in proportion to its group's size. A classifier then keeps only the synthetic rows it labels as real with high confidence. The paper uses LightGBM for this filter by default, and the survivors are pooled with the real rows 2607.19153 §3.2.5 2607.19153 §3.3.1. Card: Lazanas & Kampouropoulos (multi-generator GAN).

On the AI4I 2020 dataset, with a Random Forest, the pooled per-type generators reach PR-AUC 0.8574 and recall 0.8824 at precision 0.5696. The cost-sensitive forest reaches 0.835 and 0.5882, a single GAN over all failures 0.8245, and SMOTE 0.7525 2607.19153 tab. 4. The paper reports one run per configuration, so the 0.02 PR-AUC gap over the cost-sensitive baseline is not separated from run-to-run variation 2607.19153 §4.1. The GAN objective is the forest's validation PR-AUC, and the paper does not say whether the SVM scenarios reuse the tuned generators 2607.19153 §3.3.2. The single-GAN run took about 11 h 20 min and the multi-GAN run about 14 h 5 min, against minutes for the classical scenarios 2607.19153 §5.1.

Virtual values between samples for demand forecasting

Water-demand forecasts have large errors where demand changes sharply between two 15-minute readings 2512.06714 §1. The EDCGRU variant inserts one virtual value between each pair of consecutive readings and classifies it with the same k-means step as a real reading. The input window doubles from 96 to 192 rows 2512.06714 §3.5 2512.06714 §4.2. The authors report one virtual value per gap as the best of the values they tried, and say that more did not help 2512.06714 §4.2. Card: Salloom et al. (water demand).

The paper never gives the rule that computes a virtual value. It says only that the values relieve nonlinearity 2512.06714 §3.5, so an implementation has to choose one, such as linear interpolation, and report the choice. On two district metering areas from one data set, one-step MAPE falls from 1.26% to 0.98% and from 1.31% to 0.99% against the same model without virtual values, and training time rises from 192 s to 850 s on the first area 2512.06714 tab. 6. The abstract's 30% error reduction does not name the metric it refers to 2512.06714 §7. The k-means class features that this model also uses are an input feature, not augmentation, and are not covered here.

Language-model-written training examples (survey only)

A survey of large pre-trained language models describes two ways a model can write data for another model. The first adds labelled or silver-labelled examples to the training set, with filtering of noise. The second writes auxiliary text such as counterexamples, clarifying questions, contexts or inference rules that help another model explain or support its output 2111.01243 §5.1 2111.01243 §5.2. Card: Min et al. (pre-trained LM survey).

The survey runs no experiments and gives no result for either use, so this page can describe the method but not its effect 2111.01243 §5. For measured results with model-labelled data, see the recombined-pair and model-filtered variants above.

Token-level replacement for sequence labelling

Tagging tasks such as named-entity recognition give every token a label, so changing a word can change its label. This variant adapts four simple methods to BIO tags, and none of them needs a translation or parsing model 2010.11683 §3. Label-wise token replacement swaps a token for another token that has the same full BIO tag. Synonym replacement swaps a token for a WordNet synonym, and the extra words take I- tags when the token began a mention. Mention replacement swaps a whole mention for another mention of the same type and rewrites its tags to fit. Shuffle within segments permutes the tokens inside each mention or each run of outside tokens, and leaves the tags in place 2010.11683 §3. Label-wise replacement and shuffling keep the label sequence unchanged, while synonym and mention replacement can change it 2010.11683 tab. 1. Card: Dai & Adel (NER augmentation).

On the 50-sentence training subsets, every single method is marked significant against no augmentation, for both taggers and both data sets 2010.11683 tab. 3. For the BiLSTM tagger on MaSciP, the 50-sentence score rises from 53.0 to between 58.8 and 60.8 F1 2010.11683 tab. 3. Applying all four methods gives the largest average gain over the eight columns: +3.3 F1 for the LSTM and +1.8 for SciBERT 2010.11683 tab. 3. The gain is concentrated in the small sets. On the full training set several single-method cells are equal to or below the baseline; SciBERT on i2b2-2010 ranges from 87.2 to 87.8 against 87.8 without augmentation 2010.11683 tab. 3 2010.11683 §4.4. The released code runs only the BERT-based tagger, so the LSTM results cannot be reproduced from it (see the card's Paper vs code).

Model-filtered sentences from unlabelled abstracts

The authors of a scientific-entity tagger say that easy text augmentations risk changing entity spans, so they use a model-based procedure instead 2606.29836 §3.3. Candidate sentences come from unannotated abstracts and must contain a known entity. A tagger trained on the training set labels each candidate. A candidate is kept only when the tagger finds every known entity it matches, and the kept sentences are added until the augmented set is as large as the training set 2606.29836 §3.3. Card: Zhang et al. (NLP technology entities).

On the authors' own 311-sentence test split, the SciBERT plus BiLSTM tagger goes from 86.49 to 87.00 F1 with this augmentation 2606.29836 tab. 7. This is one model run, with no standard deviation reported 2606.29836 §4.1. The gain of 0.51 F1 is therefore not separated from run-to-run variation. The repository has no script for the procedure, so it has to be rebuilt from §3.3 2606.29836 §3.3.

This is closer to pseudo-labelling than to transforming an example, like the recombined-pair variant above. The new labels come from the model's own output, filtered by agreement with gold entities. Check the labels on a held-out split before you trust the added data.

Token labeling with MixToken for vision transformers

The class token of a vision transformer is the only token that receives the classification loss, so the patch tokens get no direct supervision 2104.10858 §1. Token labeling gives each patch token its own target. The targets come from a dense label map that an offline annotator computes once for every training image, here an ImageNet-only NFNet-F6 classifier 2104.10858 §3.2 2104.10858 §4.1. Training adds a token-level cross-entropy term to the class term 2104.10858 §3.2. MixToken is a CutMix-style mix on patch embeddings: a box of tokens is exchanged between two images, and the class target is mixed by the share of tokens from each image 2104.10858 §3.3 2104.10858 eq. 7. Card: Jiang et al. (token labeling).

The annotator is the costly step. Its top five logits per cell were stored in half precision, which the authors put at about 10 GB, against about 2 TB for full maps 2104.10858 §4.1. The paper says the added compute is negligible but gives no wall-clock time 2104.10858 §3.2.

Author-reported results: removing MixToken drops LV-ViT-S from 83.3% to 81.3% top-1 2104.10858 tab. 3. With the same ReLabel labels, MixToken scores 83.0 against 82.8 for CutMix 2104.10858 tab. 2. A weaker annotator, EfficientNet-B3 at 81.6% top-1, still helps LV-ViT-S, while NFNet-F6 gives 83.3 2104.10858 §4.2 fig. 4. The released code weights the token loss at 1.0, where the paper uses 0.5, and it does not mix token targets under MixToken 2104.10858 §3.2 2104.10858 eq. 6. Check the card's Paper vs code before copying either.

Mixup on the loss only, for segmentation

A building segmenter mixes two training images with a small weight λ, but it does not blend their label maps. It runs the network once on the blended image, then takes the cross-entropy against each label map, weighted by λ and 1 − λ 2107.12283 §6 2107.12283 eq. 5 2107.12283 eq. 6. Mixup is applied to the cross-entropy term only, not to the Tversky term 2107.12283 §6. The same pipeline uses random 448 px crops from 600 px tiles, flips, rotations, and random changes to brightness, hue, saturation and contrast 2107.12283 §6. Card: Sirko et al. (building detection).

The paper uses λ = 0.05 2107.12283 §6. With that value the partner's label map gets 95% of the loss weight 2107.12283 eq. 6. The paper does not say how partner images are chosen or how λ is sampled 2107.12283 §6. The authors kept the labels separate after preliminary runs found that blended labels did worse once pixel weights were used 2107.12283 §6. The abstract reports +0.12 mAP for mixup, while the ablation appears only as a figure 2107.12283 fig. 8. The same pipeline's self-training gives random black masks on the student's inputs, so that it learns to handle no-data regions 2107.12283 §8.

Many colour and crop views pooled over frozen features

No backbone is trained. Each slide is recoloured 50 times, and each recolouring gives 20 random square crops that a frozen ImageNet encoder embeds 1802.00752 §2.3. The haematoxylin and eosin magnitudes are scaled by two independent draws from U(0.7, 1.3) 1802.00752 §2.3. The crop features of one recolouring are pooled with a power mean (p = 3) 1802.00752 eq. 1, and a LightGBM classifier learns from the pooled descriptors 1802.00752 §2.5. Card: Rakhlin et al. (breast histology). Author-reported accuracy is 87.2% for four classes under 10-fold cross-validation on 400 images 1802.00752 tab. 1. The code also rotates crops by multiples of 90° and flips them, which the paper does not mention feature_extractor.py.

Offline rotated and noisy copies of a small photo set

A transfer-learning benchmark expanded 654 smartphone photos into 8,502 images before training. Each photo is rotated by 45, 135, 225 and 315 degrees, and each rotated copy gets noise drawn uniformly up to 10, 25 or 50 percent of the pixel range 2605.11989 §3.3.3. The paper does not say whether the noise is added to or replaces pixel values, how the corners left by rotation are filled, or whether the augmentation runs before or after the train/validation split 2605.11989 §3.3.3. If it runs first, copies of one photo can land on both sides of the split, which leaks information into validation. Card: Abou Baker et al. (transfer learning).

The results are averaged over the five datasets 2605.11989 §5.1, so the effect of this augmentation alone is not reported.

Noisy-student loop with audio augmentation

A teacher pseudo-labels one unlabelled batch at a time, and a larger student is trained on the pseudo-labelled data, the clean labelled data and augmented copies 2609.24410 §8.2 fig. 8. The augmentations are Gaussian noise, time stretch and pitch shift, applied at every generation 2609.24410 §6.2. Their magnitudes and selection are not stated, and no pseudo-label filter is described 2609.24410 §6.2 2609.24410 §7.3. Card: Safieh et al. (Jordanian speech). The authors' evaluation WER falls from 0.568 to 0.515 over five generations 2609.24410 tab. 4. Each generation also uses a larger model, and the card's results contain no run that isolates the augmentation.

Perturbed temporal graphs for reasoning fine-tuning

The reasoning model of TG-LLM is trained on verified temporal graphs, but at test time it reads graphs that its own first stage predicted. The authors treat that gap as a risk and add four perturbations to each training example: drop facts that appear in neither the question nor the chain of thought, replace relation names with synonyms, rename entities with one global map, and shift every year by one offset. The same change is applied to the question, chain of thought, candidates and answer, so the example stays consistent 2401.06853 §3.2.2 2401.06853 app. B. Card: Xiong et al. (TG-LLM).

In the authors' TGQA ablation for Llama2-13B, the augmentation raises exact match from 0.723 to 0.782 without external knowledge, and from 0.753 to 0.797 with it 2401.06853 tab. 5. These are single runs. The paper's human error table shows that not every error type falls: the temporal-graph error rate rises from 0.03 to 0.04 when augmentation is added 2401.06853 tab. 3.

The code appends one augmented copy of each training example, so the set doubles, and it draws the entity map and the year offset once per run 2401.06853 §3.2.2 SFT_TG_Reasoning.py. It sets no random seed, so two runs get different augmented data unless you fix one yourself SFT_TG_Reasoning.py. The source data is built from YAGO11k and GPT-3.5 outputs; check its terms before reusing the generated set 2401.06853 §2.

Mosaic and mixup in detector training

The YOLOv8 review names two training augmentations. Mosaic combines "four or more" images into one training sample, and mixup blends images 2408.15857 §3.1.1 2408.15857 §3.2. The paper gives no canvas size, placement rule, box update, mixup ratio or per-step probability, so neither step can be reimplemented from the paper alone 2408.15857 §3.2. It runs no experiment that isolates either one, and its claim that the augmentation helps small objects comes without figures 2408.15857 §8. Card: Yaseen (YOLOv8 review). Measure these steps on your own data before relying on them.

Cached Mosaic and MixUp, then large-scale jittering

For a real-time one-stage detector, the paper keeps a cache of decoded images and draws Mosaic and MixUp samples from it. Stage one of training uses these two steps. The last 20 epochs switch to large-scale jittering, which the paper presents as a fine-tuning stage close to the real image distribution 2212.07784 §3.3 2212.07784 tab. 1. A large cache with random eviction behaves like uncached augmentation. A small cache with first-in, first-out eviction makes the same image reappear with different augmentations in one batch 2212.07784 §3.3. Cached Mosaic takes 24.0 ms per 100 images against 87.1 ms uncached 2212.07784 tab. 7. Card: Lyu et al. (RTMDet).

On COCO, the two-stage schedule with cached Mosaic and MixUp reaches 51.3 AP for RTMDet-l, against 46.7 with large-scale jittering alone and 49.8 with uncached Mosaic and MixUp in both stages 2212.07784 tab. 7. The ablations are single runs with no spread reported. The paper's numbers are for its own detector family, so they do not transfer directly to other models.

Flips and rotations, split before or after

A grain-variety classifier grows about 4,500 captured images to 27,000 with horizontal and vertical flips and rotations of 45, 90 and 180 degrees 2609.10524 §4.1 2609.10524 fig. 2. The paper does not say whether this happens before or after the train/test split. Table 2 gives 22,500 training and 4,500 test images, and the test count equals the per-class count before augmentation. That suggests a split on originals, but the paper does not say so 2609.10524 tab. 2. If the split came after augmentation, copies of one grain could sit on both sides, so test accuracy would overstate the result 2609.10524 §4.1. The reported base-model accuracies of 97.5 to 99.2 percent are validation figures, and the results tables give no run without augmentation 2609.10524 tab. 4. Card: Islam et al. (rice variety).

Heavy transform chain for a fine-tuned classifier

Two ImageNet-pretrained networks, ResNet-50 and Inception-v3, are fine-tuned on field photos of eight weed species and a Negatives class 1810.05726 § Deep Learning. Each training image is resized to 256 × 256 and then rotated over the full circle, zoomed by a factor between 0.5 and 1, sheared, shifted per colour channel by ±25 on the 8-bit scale, scaled in brightness between 0.75 and 1.25, and flipped with probability 0.5. A centre crop to 224 × 224 follows 1810.05726 § Deep Learning deepweeds.py. Validation images go through the same chain deepweeds.py, so the validation loss that drives the learning rate and early stopping is computed on randomly altered images deepweeds.py. Card: Olsen et al. (DeepWeeds).

The paper describes random perspective transforms and a separate intensity shift of ±25 on all channels. The released code has shear and a brightness factor instead, with no perspective transform and no separate intensity shift 1810.05726 § Deep Learning deepweeds.py. The text does not say which chain produced the reported results.

The paper says that training without augmentation over-fitted strongly, but it gives no numbers for that 1810.05726 § Deep Learning. The reported class-weighted top-1 accuracy over the five test folds is 95.7% for ResNet-50 and 95.1% for Inception-v3 1810.05726 tab. 2. No run without augmentation is reported, so the effect of the chain alone is not measured.

Random tone mapping for lighting-sensitive depth

This variant changes only the brightness of the RGB image. The depth map is kept as it is, because a change of light does not move a surface. Each training image gets a gamma curve with γ = 1/2.2. The image is then scaled so that a chosen percentile of its intensity maps to 0.8, and the percentile is drawn between 70 and 99 2408.06083 §3.1 2408.06083 eq. 1. The paper states the scale only in words, and it does not say how the percentile is sampled, whether it is drawn per sample or per iteration, or what intensity range the input has 2408.06083 §3.1. Card: Zhang et al. (non-Lambertian depth).

The motivation is that glass and mirror surfaces change in appearance under light. Strong light can wash out glass texture, and weak light adds noise and colour drift 2408.06083 §1. Applying the curve needs the original scene assets, which the paper says exist for its synthetic training set 2408.06083 §3.1.

On the Booster region with ToM surfaces, δ1.05 rises from 0.644 for the Depth Anything V2 baseline to 0.800 with this augmentation alone. The guidance loss alone reaches 0.815, and the two together reach 0.859 2408.06083 tab. 3. The baseline row of that ablation equals the public model's row in the main table, and the paper does not say whether it was retrained on the same data without this augmentation. The size of the augmentation's own effect therefore rests on that unstated choice 2408.06083 §4.4.1. The numbers are author-reported on the Booster split with ground truth 2408.06083 §4.1.

Library transform pipeline with GPU batch augmentation

The fastai library builds a training pipeline from blocks. Per-item transforms run on the CPU, and batch transforms run on the mini-batch, on the GPU when one is available 2002.04688 §2.1. The Pets example uses a batch transform set with a warp of zero, and it leaves rotation, zoom and lighting at their defaults 2002.04688 §2.1. The paper gives no accuracy effect for augmentation alone. Its defaults are a moving target: the paper describes v2.0.0, and the defaults at the pinned code commit differ 2002.04688 §1. Card: Howard & Gugger (fastai).

Tabular CutMix and mixup for contrastive pre-training

SAINT is a transformer over the cells of a table row. It can be pre-trained on unlabelled rows, where each row has two views 2106.01342 §4. The second view applies CutMix in raw input space: each cell may be replaced by the same cell from a partner row, chosen by a binary mask 2106.01342 eq. 3. The embeddings of that view are then mixed with the embedding of another row, with weight α 2106.01342 eq. 4. An InfoNCE loss pulls the two views of a row together, and a denoising loss rebuilds the original cells 2106.01342 §4 2106.01342 eq. 5. Card: Somepalli et al. (SAINT).

In the paper's ablation with 50 labelled rows, the mean AUC is 86.96 with both augmentations, 85.37 with CutMix alone and 86.45 with mixup alone 2106.01342 tab. 9. Pre-training gives 86.91 against 85.78 without it at 50 labels, and 92.99 against 93.13 when all labels are used 2106.01342 tab. 3. The gain is therefore limited to few labels. All figures are author-reported and averaged over the 14 binary datasets.

The released code differs from the paper in its defaults. It turns augmentation off by default, uses 0.1 as the CutMix swap probability and 0.3 as the mixup weight train.py train.py train.py. The paper gives p = 0.3 for the CutMix mask without saying whether it marks cells to keep or to swap, so the swap rate is not settled by the paper 2106.01342 §5 2106.01342 eq. 3. Predictions also depend on the other rows in the batch, because intersample attention runs across the batch model.py.

Resampling for class imbalance

A benchmark pairs nine resampling methods with nine ensemble models on imbalanced tabular sets 2304.02858 §4.1 2304.02858 tab. 1. The resamplers are SMOTE and four of its variants (SMOTE-ENN, Borderline-SMOTE, SVM-SMOTE and K-means SMOTE), ADASYN, random over- and under-sampling, and CT-GAN, which is a GAN 2304.02858 tab. 1 2304.02858 §3.1 2304.02858 §3.3. The resampler acts on the training rows only, and the classifier is unchanged 2304.02858 §5.1. Each pair is run 30 times on a 60:40 split 2304.02858 §5.1. Card: Khan et al. (imbalance benchmark).

The conclusion names SMOTE with LightGBM and ROS with LightGBM as the best pairs, and says that plain SMOTE and ROS beat CT-GAN at lower cost 2304.02858 §7. The per-dataset best-pair table does not match that: CT-GAN is the best pair on five of its rows 2304.02858 tab. 8. The paper reports no timing or memory figures, so the cost claim has no data behind it 2304.02858 §7. Read the tables rather than the conclusion.

Three points from the released code limit how the numbers can be read. The loop resamples the output of the previous run, not the original training split, so run-to-run variation differs from the protocol the paper describes main.py. The AUC column is computed from hard predictions, which makes it balanced accuracy rather than a threshold-free AUC 2304.02858 eq. 21 main.py. The split is not stratified, so a test set on a rare class can hold only a few minority rows 2304.02858 §5.1 2304.02858 tab. 2.

Learned augmentation policies (survey only)

A survey of few-shot learning sorts augmentation into hand-crafted rules and learned processing 2205.06743 §3. Hand-crafted rules include random erasing, cutout-style filling, random cropping and mixing, and feature-level variants 2205.06743 §3.1. Learned processing uses meta-learning to choose or generate augmentations. The survey names an augmentation-policy search (DADA), conditional generators (AFHN, MetaGAN) and a delta-encoder 2205.06743 §3.2. It runs no experiment of its own, and its tables mix protocols, so they cannot rank these methods against each other 2205.06743 tab. 9 2205.06743 tab. 12. Card: Song et al. (few-shot survey). The method is described here for orientation only; read the original papers before implementing it.

Text augmentation taxonomy (survey only)

A survey sorts text augmentation into paraphrasing, noising and sampling, ordered by how much new diversity each adds 2110.01852 §2. Its properties table marks Mixup as the only online method, and most non-learnable methods as dependent on external resources such as WordNet [arxiv:2110.01852 §2.4] tab. 1. Its examples include EDA-style swaps, deletions and insertions for noising 2110.01852 §2.2, and word-piece replacement with probability 0.4 2110.01852 §2.1.3. The survey runs no experiments of its own, so it gives no effect sizes 2110.01852 §4. Card: Li, Hou & Che (NLP augmentation survey).

Input-side regularizers in a CNN survey (survey only)

A survey of recent CNN regularizers files Cutout, RandomErasing, AutoAugment, PBA, RandAugment, Mixup and CutMix under input methods 2201.03299 §1.3 2201.03299 §3. It reports that about 44% of the works it surveys change the input 2201.03299 §1.5. It calls RandAugment the likely best input method, which is the survey's opinion and was not tested 2201.03299 §7.1. The Cutout patch side it lists is 16×16 for CIFAR-10 and 8×8 for CIFAR-100 2201.03299 §3.1. Its compiled numbers are the original authors', not reproduced 2201.03299 tab. 2. Card: dos Santos & Papa (CNN regularization survey).

Manipulation, erasing and mixing measured across three tasks

A survey of image augmentation that also runs its own with and without comparisons for the basic methods: manipulation, erasing and mixing 2204.08610 §4. Printed changes are +0.53 to +2.61 mean IoU for four segmenters on PASCAL VOC 2204.08610 tab. 2, +0.07 to +2.80 accuracy for three networks on CIFAR-10, CIFAR-100 and SVHN 2204.08610 tab. 3, and, on COCO2017, +0.40 mAP for Faster R-CNN but −0.27 for CenterNet 2204.08610 tab. 4. Training settings are not stated and no code is released 2204.08610 §4. Card: Yang et al. (image augmentation survey).

Physics-rendered overlays on real photos

Paired training data for raindrop removal is hard to collect, so the authors build it. They start from clean Cityscapes street photos and place virtual raindrops on them. Each drop is a spherical cap in front of the camera. A ray through a drop is refracted, and the pixel takes the background value where the refracted ray meets the background plane. Drop regions are then blurred with a disk kernel of 7 to 20 pixels. The drop positions come from the simulation, so the binary drop mask comes with each image 2609.37870 §3. Each background gets 50 to 70 drops. The training split has about 30,000 images from the Cityscapes training set, and the test split has 1,525 rainy and clear pairs from the Cityscapes test set 2609.37870 §3 2609.37870 §5.1.

The clean photo is the target and the rendered photo is the input, so the label stays valid by construction 2609.37870 §3. The base images are real and labelled, which separates this from the 3D-scan case in the scope section below. The rendering stage is under-specified. The sampling distributions, the refraction equations and whether drop parameters are drawn per drop or per image are not stated. No renderer or dataset release is mentioned 2609.37870 §3. Card: Hao et al. (raindrop removal).

Stitched face forgeries with known masks for deepfake detection

This variant makes forged training frames rather than transforming real ones. A real target frame is paired with a frame of another identity whose 68 facial landmarks are close to its own. The source face is aligned to the target, and it is blended in through a mask. The mask starts as the convex hull of the target landmarks, is deformed elastically, and is then blurred 2012.09311 §3.2 2012.09311 alg. 1. The label is valid by construction, because the method knows which pixels it pasted. Each real sample in training is replaced by one of these forgeries with probability 0.5 2012.09311 §4.2. The label is a pairwise consistency map and a real/fake flag, not a class 2012.09311 §3.1. Card: Zhao et al. (self-consistency deepfake).

The paper's ablation trains the same detector with and without these forgeries. With DF frames as the real source, the four-set average AUC is 79.19 without them and 86.37 with them. With DFDC-P as the source, it is 74.78 and 86.35 2012.09311 tab. 8. The cross-dataset numbers train on real FF++ frames plus these forgeries, and the in-dataset numbers use them by default, so the two protocols are not directly comparable 2012.09311 §4.2.

The blending method is picked at random from a set the paper does not list, and the landmark distance threshold and the augmentation strengths are not stated 2012.09311 §3.2 2012.09311 alg. 1. The released code is not public, so these values have to be chosen again for a reimplementation. The numbers are author-reported 2012.09311 §4.3.

Synthetic video from annotated still images

Labelled video for saliency is scarce, so the authors turn each annotated still image into a training pair. They generate ten simulated frames per image and use the warp I'(x) = I(x + v(x)), where v is a flow field 1702.00871 §4 1702.00871 eq. 5. The field is built on superpixels. Background parts are split into seeds that move at random and the rest, and foreground parts share one main motion with small variation 1702.00871 §4. A smoothing energy with a unary and a neighbour term then keeps the seeds and the foreground near their initial motion 1702.00871 eq. 6 1702.00871 eq. 7 1702.00871 eq. 8. The authors report more than 150K synthetic clips, generated in about three hours 1702.00871 §4 1702.00871 §5.1. The mask is warped with the same field, so the label moves with the pixels 1702.00871 §4.

The gain comes from mixing this data with real video. On the FBMS test set, the full model reaches 7.65% MAE. Training on image-derived data alone gives 9.27%, and training on video alone, which is the smallest set, gives 24.5% 1702.00871 tab. 2 1702.00871 §5.3.

The paper does not state how pixels that leave the frame or become occluded are filled, which interpolation the mask warp uses, or how many superpixels are made 1702.00871 §4. The horizontal motion bound is also not stated 1702.00871 §4. Reimplementers must choose these. Card: Wang et al. (video saliency).

Cross-painting robot demonstrations onto other arms

Robot demonstrations come mostly from a few arms, so a policy often works only on the arm it saw. This variant reuses each recorded trajectory and renders it with another robot. For each frame, a fused mask finds the source arm: a learned segmenter, fine-tuned on a few labelled trajectories per dataset, and a simulation mask, aligned to the image by a small grid search. A video inpainting model fills the background behind the arm, and the target arm is replayed in simulation along the same trajectory and composited over that background 2512.13100 §4.1 2512.13100 §4.2. Each target base position is tuned in simulation so that every pose is reachable, and a trajectory with no feasible base is discarded 2512.13100 §4.2 (2). Card: Ji et al. (OXE-AugE).

The authors applied this to 16 Open X-Embodiment datasets, with eight or nine target robots per dataset, and report more than 4.4M augmented trajectories 2512.13100 §6 2512.13100 app. A.2 2512.13100 tab. 1. Two datasets get no WidowX augmentation, because most of their trajectories exceed that arm's workspace 2512.13100 app. A.2.

The authors report results from two setups. In simulation, with a Diffusion Policy on five Robosuite tasks, adding one augmented target to the source demonstrations improves success on that target by 9 to 19%, against the augmented target data alone. Augmented training also makes the source robot more robust under lighting and occlusion changes 2512.13100 §5.1 2512.13100 fig. 3 2512.13100 fig. 4. In the real tests, OpenVLA-OFT and pi0 fine-tuned with augmented Bridge data gain 24% and 45% on average, against fine-tuning on Bridge alone, on unseen arm and gripper pairs 2512.13100 §6.2. The text gives these gains without a unit or denominator, and the real-test score is a partial credit, not a success rate 2512.13100 app. A.5.

The paper also compares this with diffusion models trained to translate robot appearance on 700K paired images. Fine-tuning on those images lowers final policy performance by 27 to 30%, in the authors' setup 2512.13100 §6.2. The method assumes that grippers are similar in shape and function, and that all frame transforms are known 2512.13100 §3. Use it where those hold and where a simulator of the target arm is available.

Route-and-instruction pairs written by a multimodal model

Navigation agents that follow spoken directions learn from pairs of a route and a sentence that describes it. This variant builds the pairs from ordinary house-tour videos instead of a simulator. Each video is sampled as a path of room nodes and transition nodes. CLIP labels each room with its type and salient objects, and an action inverse model estimates the action between consecutive rooms 2411.11394 §3.1. A multimodal model then receives the room images, labels and actions in order, and writes an instruction for the path 2411.11394 §3.2. Card: Yan et al. (InstruGen).

The check is what keeps the label valid. The model is asked to extract the room and action pairs from its own instruction. If they differ from the route's labels, the instruction is regenerated, and regular expressions remove stray characters and prompt fragments 2411.11394 §3.2. The check compares the instruction with the route's labels, so an error in CLIP or in the action model passes through it unless it breaks the instruction's logic 2411.11394 §3.1 2411.11394 §3.2. The mismatch test, the retry limit and the number of accepted pairs are not stated.

The ablation on R2R val-unseen adds the steps one at a time. Without any check, the success rate is 67.42. Adding the consistency check gives 69.32, the regex step 70.90, and a manual review of the best row 71.10 2411.11394 tab. 1. The last step is the only one that is not automatic, and the paper does not say how much review was done. The automatic-only row is therefore the figure to compare when no one can review the pairs.

The prompt can ask for a concise or a detailed instruction 2411.11394 §3.2. In the granularity table, coarse-grained pretraining beats fine-grained pretraining for every fine-tuning set. The best row pairs coarse-grained pretraining with the R2R* fine-tuning set, at 71.10 success on R2R val-unseen 2411.11394 tab. 2. The card does not establish how the coarse-grained set maps onto the concise and detailed prompts. R2R* mixes the generated instructions with the original R2R ones, and the mix ratio is not stated 2411.11394 §4.2. The paper's comparison table places this above Lily, at 70.00 2411.11394 tab. 3. The one-shot result, with one training environment averaged over five random sets, is 51.25 against 50.86 for Lily, a margin under one point with no spread reported 2411.11394 tab. 4. The RxR zero-shot score is 29.35, with a model trained on R2R* only 2411.11394 tab. 5. The comparison table does not say whether its baseline rows were re-run or copied 2411.11394 tab. 3.

The released repository holds only a README, so the prompts, the regex rules and the generated set cannot be checked. The prompt appears only as a figure, and the learning rate, the optimizer and the pretext-loss weights are not stated 2411.11394 fig. 2 2411.11394 §4.1. The generated set's size is not given beyond the 3,879 source videos 2411.11394 §3.1. The model's version, temperature and sampling settings are not stated 2411.11394 §3.2.

Laser-beam band swap between scans for semi-supervised LiDAR segmentation

Each scan is split into areas by laser inclination. Two scans then exchange alternating areas, so that every area's neighbours come from the other scan 2405.05258 §3.2 eq. 7. A teacher, whose weights are an exponential moving average of the student's, pseudo-labels the unlabelled scans, and only predictions above a confidence threshold are kept 2405.05258 §3.3. LaserMix++ adds camera features: each point is paired with its image pixel, the number of areas is drawn from 2 to 6, and point features are distilled towards image features 2405.05258 §4.2 2405.05258 §4.3. Card: Kong et al. (LaserMix++). The authors report an average gain of 2% to 3% for LaserMix++ over LaserMix 2405.05258 §5.3, and a gain of 3.1 to 5.6 mIoU with Cylinder3D at 1% labels 2405.05258 tab. 2. The linked repository holds the LaserMix code only, so the camera parts are not reproducible from it README.md, and several released config values differ from the paper, including the pseudo-label threshold.

Object-first LiDAR generation for detector augmentation

Real LiDAR objects are sparse and partial, and foreground points are a small share of a scene's points. The authors say that a scene-level generator therefore spends most of its capacity on background 2412.17226 §3.2. This variant generates each foreground object first, from a text description and a 3D box 2412.17226 §3.2. Each object comes from a point-cloud diffusion model that takes a CLIP text embedding and a Fourier embedding of the box as its condition 2412.17226 eq. 1 2412.17226 eq. 2 2412.17226 eq. 3. The objects are then projected into a range image, and a scene range-image diffusion model receives them through a ControlNet-style controller 2412.17226 §3.2 2412.17226 eq. 7. Card: Yan et al. (OLiDM).

An extra per-category loss, called object semantic alignment, splits the scene range image into one channel per category and adds a second noise-prediction loss for it 2412.17226 §3.3 2412.17226 eq. 9. The paper adds this to the scene loss, but it does not state the weighting 2412.17226 §3.3. The method assumes that each training object has a category, a caption and a 3D box 2412.17226 §3.2 2412.17226 app. A. Captions come from BLIP-2 2412.17226 app. A.

Scene-level generation is compared on KITTI-360. Against UltraLiDAR, the paper reports FPD 7.60 against 25.1 and JSD 0.42 against 0.71 2412.17226 tab. 1. Removing the per-category loss raises FPD to 9.20 2412.17226 tab. 1, so the paper credits that loss with part of the gain. Foreground quality is reported in a separate table, where the full model has the lowest Chamfer distance, 1.88, against 17.6 for R2DM 2412.17226 tab. 2.

The downstream claim is about detection. Generated objects are placed on the ground plane, uniformly, to avoid oversampling near the ego vehicle 2412.17226 app. E. On nuScenes, PointPillars goes from 44.8 mAP with no augmentation to 46.7 with OLiDM objects, against 45.1 with ground-truth copy-paste augmentation. CenterPoint goes from 59.0 to 61.9 2412.17226 tab. 4. The number of objects added is not stated 2412.17226 app. E. The paper reports no standard deviation or seed 2412.17226 §4.1.

Training costs are reported in the appendix. OLiDM takes 19 hours of training on the authors' KITTI-360 setup, against 12 hours for R2DM, and 3.5 hours of inference 2412.17226 app. C. The cost table gives no hardware and no sample count for the inference figure. Scene generation runs 300k steps at batch 16 2412.17226 app. A.

Eq. 5 as printed does not define an attention operation on its own, and eq. 7 and eq. 8 disagree on what the controller receives. The paper does not say which version was trained 2412.17226 eq. 5 2412.17226 eq. 7 2412.17226 eq. 8. Plan the object denoiser's attention yourself, and check the released code before reproducing the numbers. The released repository has no model code, so there is nothing to check against yet.

3D rotation, scaling and shear for skeleton sequences

Skeleton action recognition takes 3D joint coordinates as input, so image transforms do not apply directly. The two-stream RNN perturbs those coordinates during training only 1704.02581 §4.3. Three transforms are used. The rotation is composed from three axis rotations, and the input is rotated at random "within a certain range" about the x and y axes. The authors tie this to viewpoint change, with the camera plane perpendicular to z 1704.02581 §4.3 1704.02581 eq. 5. Scaling is a diagonal matrix with scale factors for each axis, drawn per sample to cover different body sizes 1704.02581 eq. 6 1704.02581 §4.3. Shear is a 3×3 matrix with six off-diagonal factors 1704.02581 eq. 7. Card: Wang & Wang (two-stream skeleton RNN).

The paper gives no ranges, probabilities or copy counts for any of the three 1704.02581 §4.3. The text applies the rotation to x and y, but the composite in eq. 5 has three axes, and whether a z rotation is drawn is not stated 1704.02581 eq. 5. The card's search found no official code, so an implementer has to choose the ranges.

In the authors' NTU ablation, the two-stream model without a transform reaches 68.6% cross-subject and 71.7% cross-view. With the 3D transform it reaches 71.3% and 79.5% 1704.02581 tab. 1. On SBU the same change moves 91.9% to 94.8% 1704.02581 tab. 1. The table row does not say which of the three transforms it includes, so the ablation cannot separate them 1704.02581 tab. 1. Use this where a change of camera angle or body size should leave the action label unchanged.

When to use which

Implementation sketch

The Copy-Paste step from 2012.07177 §3 in Python-flavoured pseudocode. Both images are assumed to be resized to one canvas size before compositing; the paper pads small images with gray 2012.07177 §3.

# Copy-Paste for one training example  [arxiv:2012.07177 §3]
# Input:  labelled set D with masks, boxes and classes per instance
# Output: composite image and its updated instance list

def copy_paste_example(D):
    # Sample two training images at random  [arxiv:2012.07177 §3]
    main, paste = sample(D, 2)
    # Scale-jitter and flip each image on its own  [arxiv:2012.07177 §3]
    main = scale_jitter_and_flip(main)
    paste = scale_jitter_and_flip(paste)
    # Choose a random subset of the pasted image's objects  [arxiv:2012.07177 tab. 10]
    chosen = random_subset(paste.instances)
    # Union of the chosen masks, binary, shape (H, W)  [arxiv:2012.07177 §3]
    alpha = union([inst.mask for inst in chosen])
    # Composite the pasted pixels over the main image  [arxiv:2012.07177 §3]
    image = paste.image * alpha + main.image * (1 - alpha)
    # Trim or drop main-image objects covered by the paste  [arxiv:2012.07177 §3]
    kept = []
    for inst in main.instances:
        mask = inst.mask * (1 - alpha)
        if mask.sum() == 0:
            # Fully covered: remove from the labels
            continue
        # Partly covered: keep, with box recomputed from the new mask
        kept.append(inst.with_mask(mask))
    return image, kept + chosen

Where the variants differ: the graph variant replaces the composite step with feature masking or edge edits applied to two copies of a graph. The label-preserving variants replace it with a fixed set of flips, colour inversions or generic geometric transforms 2103.00111 §6.1 1802.09567 §4.2 1811.04241 §4. The saliency variant keeps the paste step but chooses the patch from a saliency peak and mixes labels by patch area 2006.01791 §3.2. The feature-space variant skips images entirely and samples vectors from a Gaussian per support feature 2101.06395 §3.2.3. The repeated-copy variant keeps the image pipeline and changes only how many times each scarce example is drawn per step 2201.09206 §3.3. The sentence-pair variant samples pairs instead of transforming them, and the generative variant replaces the input image with a class-name prompt 2010.08240 §3.1 2304.08466 §4.1. The temporal-graph variant changes graph facts, relation names, entity names and years together, using the same mapping for every field 2401.06853 app. B. The time-series variant makes two views of one series, one weak and one strong, and the model-filtered variant samples new sentences that a tagger has checked 2106.14112 §3.1 2606.29836 §3.3. The per-type GAN variant replaces the transform with one generator per failure type and pools its filtered samples 2607.19153 §3.2.5. The virtual-value variant inserts a value between each pair of readings, with the rule left to the implementer 2512.06714 §3.5. The heavy-chain variant is the standard per-image pipeline with a fixed order of steps deepweeds.py. The tabular variant swaps cells between rows and mixes embeddings 2106.01342 §4. The resampling variant adds minority rows before fitting and leaves the classifier unchanged 2304.02858 §4.1. The learned edge-drop variant replaces the fixed drop rule with a probability per edge, produced by a network that is trained adversarially against the encoder [arxiv:2106.05819 §3.2]. The synthetic-video variant warps a still image along a flow field to make the second frame of a training pair 1702.00871 §4. The robot cross-painting variant replaces the image transform with a segment, inpaint and replay pipeline, and drops trajectories with no feasible target base 2512.13100 §4.1 2512.13100 §4.2. The route-instruction variant replaces the image step with a path sampler, a labeller and a checked text generator 2411.11394 §3.1 2411.11394 §3.2. The LiDAR variant generates objects first and then places them in a range-image scene model 2412.17226 §3.2. The skeleton variant perturbs the joint coordinates with a random 3D rotation, scaling and shear before either stream reads them 1704.02581 §4.3.

Contradictions

• Random patch placement. Copy-Paste pastes objects at random positions and reports gains 2012.07177 §3. SaliencyMix argues that a random patch can land on background and teach the model that background belongs to the source class 2006.01791 §1. The tasks differ (instance segmentation against classification), so the two claims are not directly comparable.
• SaliencyMix claims against its tables. The text calls its CIFAR-10 result of 2.76% the best known top-1 error, and says it beats all other methods except PuzzleMix on CIFAR 2006.01791 §4.1.1. Table 1 has AutoAugment at 2.60% on that column 2006.01791 tab. 1. On ImageNet, the text says SaliencyMix beats all other methods, but table 2 has PuzzleMix ahead on top-1 (21.24 against 21.26) and top-5 (5.71 against 5.76) 2006.01791 §4.1.2 2006.01791 tab. 2. Use the tables, not the text.
• Source of the ALPR gain. The card for 1802.09567 reports that augmentation is essential for letter recognition, while the paper's own discussion attributes the main gain to splitting letters and digits into two networks, with no ablation separating the two 1802.09567 §5.1.4. Treat the size of the augmentation effect as unmeasured.
• Graph augmentation numbers. The survey's accuracy table combines results from different papers and protocols, and it does not say which values were re-run 2103.00111 §8. Gaps between methods on that table are not a controlled comparison of augmentations.
• Which baseline row is the k = 1 model (FSRA). The k = 1 row of the main comparison matches the row with both the KL loss and the triplet loss in the ablation, while the text says the KL loss was not used in the final model 2201.09206 tab. 2 2201.09206 tab. 7 2201.09206 §3.4. The triplet-only row differs by 0.03 AP, so the size of the copy effect is not affected, but the k = 1 baseline is ambiguous.
• Fixed against learned graph drops. Both the centrality-based method and the learned method report gains over uniform or non-adversarial drops, but on different tasks and protocols. The centrality method is evaluated on five node classification datasets 2010.14945 tab. 4. The learned method is evaluated on graph classification with linear probes. Its paper says an SVM protocol makes untrained encoders look much better, which is why it avoids that protocol 2106.05819 §5.1. The two sets of numbers cannot be ranked against each other.
• Multi-GAN F1 in the text against its table. The text calls the multi-GAN F1 of 0.6923 the highest of all conditions, but the baseline forest's F1 in tab. 4 is 0.740 2607.19153 §4.1 2607.19153 tab. 4. Use the table, not the sentence.
• CT-GAN in the imbalance benchmark. The conclusion says CT-GAN is not a top performer 2304.02858 §7. The best-pair table names CT-GAN best on five rows 2304.02858 tab. 8. Both cannot hold; use the table.
• Raindrop removal claims against its tables. The introduction claims a clear lead over every existing method on all existing datasets and on real rainy images 2609.37870 §1. On the real-image set of Qian et al., Table 2 has Qian et al. ahead on PSNR, 30.82 dB against 30.17 dB for the rough-mask model. The rough-mask model has the higher SSIM, 0.9128 against 0.9050 2609.37870 tab. 2. The text calls the result "on par" 2609.37870 §5.2. Use the table, not the introduction.
• Cached-Mosaic gain for small models. The text says a small FIFO cache helps the small models by about 0.5 AP 2212.07784 §4.4. Table 7 gives 44.2 against 43.9 for RTMDet-s, a gain of 0.3, and 51.1 against 51.3 for RTMDet-l, a loss of 0.2 2212.07784 tab. 7. Use the table.
• Mixing count in RTMDet. The text says eight images are mixed per sample 2212.07784 §3.3. The official configs combine four images in Mosaic and two in MixUp rtmdet_l_8xb32-300e_coco.py. Treat the count as the config's, not the text's.
• Adaptive gain on Amazon-Computers. The text says the full adaptive scheme gives 1.5 points over the base model 2010.14945 §4.3. Table 4 gives 87.85 against 86.25, a difference of 1.60 2010.14945 tab. 4.
• Printed changes against printed values (image survey). The "reported change" for DeepLabV3+ is +0.53, but its printed values 75.32 and 75.75 differ by 0.43 2204.08610 tab. 2. The SVHN Wide-ResNet row prints +0.80, but its values differ by 0.08 2204.08610 tab. 3. For Faster R-CNN, the mAP change printed as 2.40 is 0.40 from the printed values [arxiv:2204.08610 tab. 4]. Use the values, not the printed change column.
• Relative WER gain (speech). The text calls the gain an "8% relative improvement", but the card's computation from Table 4 gives 9.3% 2609.24410 §9 2609.24410 tab. 4.
• Detection gains in the OLiDM paper against its table. The headline gives mAP +2.4 and NDS +1.9 on nuScenes 2412.17226 abstract. The mAP figure is the mean of the two mAP gains in the table, but the NDS gains in tab. 4 are +2.8 for PointPillars and +2.7 for CenterPoint, so +1.9 does not match. The table also labels the GT-Aug row "Real" for PointPillars and "Synthetic" for CenterPoint 2412.17226 tab. 4. Use the table, and treat the grouping of GT-Aug as unresolved.

Scope: what is not augmentation

Some steps in the cited papers add or select training data without transforming existing examples. They are not variants on this page.

Omnidata renders new views and labels from textured 3D scans, so it creates labelled examples rather than transforming existing ones 2110.04994 §3.1. It is a dataset-generation method, and it is not one of the variants above; see the dataset-curation concept for dataset construction. Card: Eftekhar et al. (Omnidata). The released starter data and weights are licensed for non-commercial research only, which matters for any use of its outputs; the card's Pitfalls section has the details.

Test-time augmentation is also out of scope for training. It averages a classifier's outputs over transformed copies of one input, and the weights for combining them are learned on held-out data 2011.11156 §4. The paper reports that plain averaging can turn correct predictions wrong 2011.11156 §3.2. Card: Shanmugam et al. (TTA aggregation).

Random-view scoring for computer-aided detection is also test-time averaging. Each candidate is cut into random translated and rotated cubes at several scales, and the ConvNet's probabilities are averaged over them 1505.03046 §II-C 1505.03046 eq. 1. The aggregation is applied to the views at scoring time, so it belongs with the test-time entry above. Card: Roth et al. (random view aggregation).

Some cards name augmentation without describing it, so they give nothing to build a variant from. The drug-discovery review proposes synthetic data for scarce data but describes no method and reports no experiment 2212.08104 §7. The emotion-detection paper says that data augmentation, oversampling and re-weighting are used, but states no factor, method or class weight 2605.30099 §IV.B. Cards: Blanco-Gonzalez et al. (AI in drug discovery review) and Teye et al. (emotion detection).

The ensemble in 1804.02386 §3.2.6 trains seven copies of one network, each on a bootstrap sample of the training segments drawn with replacement, and averages their outputs. This is resampling for an ensemble, not a transform of examples. The card carries this concept tag, but it describes no augmentation of trajectories, so the paper is listed here for scope only. Card: Dabiri & Heaslip (GPS transport modes). Its ensemble moves test accuracy from 82.3% to 84.8% 1804.02386 tab. 4. The table has no row with one bootstrap copy alone, so that number cannot be credited to resampling rather than to averaging.

Papers

• Ghiasi et al. (Copy-Paste): pastes random instance subsets between two jittered training images, with mask and box updates for covered objects, for instance segmentation.
• Laroca et al. (ALPR): colour-inverted and flip-invariant character samples for a plate recogniser, with the valid flips listed per dataset.
• Liu et al. (graph SSL survey): taxonomy of graph pretext tasks, including feature- and topology-level augmentations as contrastive views.
• Alom et al. (IRRCNN): generic rotation, shift, shear, zoom and flip chain for histopathology, with a with and without comparison.
• Chowdary et al. (face mask): eight training-image operations for transfer-learned InceptionV3 on a simulated mask dataset.
• Yang et al. (distribution calibration): training-free feature sampling from borrowed Gaussians for few-shot classification on frozen features.
• Dai et al. (FSRA): repeated augmented satellite copies per step for drone-to-satellite geo-localisation.
• Park & Kim (ViT analysis): analysis of vision transformers, with the strong augmentation recipe used in its experiments.
• Eftekhar et al. (Omnidata): synthetic multi-task dataset rendered from 3D scans; listed for scope, not as an augmentation method.
• Azimi et al. (steel microstructure): rotated 1000 × 1000 px patches with class-balanced strides for pixel-wise steel microstructure segmentation.
• Dabiri & Heaslip (GPS transport modes): seven-model bootstrap ensemble for GPS mode inference; listed for scope, not as an augmentation method.
• Uddin et al. (SaliencyMix): pastes a patch centred on the most salient pixel of one image onto another and mixes labels by patch area, for CNN classification.
• Thakur et al. (Augmented SBERT): recombines gold sentences into new pairs, labels them with a cross-encoder, and trains a bi-encoder on the result.
• Azizi et al. (synthetic ImageNet): fine-tuned text-to-image generator with class-name prompts, used to add synthetic images to ImageNet training.
• Xiong et al. (TG-LLM): four perturbations of synthetic temporal graphs, applied with the same entity and year maps to every field of a training example.
• Yaseen (YOLOv8 review): names mosaic and mixup as YOLOv8 training steps; a review with no experiments of its own.
• Islam et al. (rice variety): flips and rotations that grow a grain photo set sixfold before training ten CNNs.
• Dai & Adel (NER augmentation): four token-level methods for BIO-tagged sequences that keep tags valid.
• Jiang et al. (token labeling): MixToken, a CutMix-style mix on patch embeddings with dense token targets.
• Sirko et al. (building detection): mixup on the segmentation loss only, with separate label maps.
• Abou Baker et al. (transfer learning): offline rotated and noisy copies of a small smartphone photo set.
• Shanmugam et al. (TTA aggregation): learned weights over test-time augmented views; listed for scope, not as training augmentation.
• Kamilaris & Prenafeta-Boldú (agriculture survey): survey of deep-learning studies in agriculture, with the augmentations they used tallied; no experiments of its own.
• Eldele et al. (TS-TCC): weak and strong perturbations of time series as the two views of a contrastive pretraining step.
• Zhang et al. (NLP technology entities): tagger-filtered sentences from unlabelled abstracts for scientific-entity recognition.
• Shen et al. (mammography): label-safe random transforms drawn on the fly for whole screening mammograms.
• Abou Baker et al. (ewaste): reflections, shifts, scales and shears that expand a small smartphone-photo set nine-fold before the split.
• Min et al. (pre-trained LM survey): survey of pre-trained language models, including model-written training data; no experiments of its own.
• Salloom et al. (water demand): one virtual value between consecutive demand readings for GRU forecasting, with k-means class features.
• Lazanas & Kampouropoulos (multi-generator GAN): one WGAN-GP per failure type, with pooled synthetic failures for tabular predictive maintenance.
• Roth et al. (random view aggregation): random cubes around each detection candidate, scored by a ConvNet and averaged; listed for scope, not as training augmentation.
• Olsen et al. (DeepWeeds): ImageNet-pretrained ResNet-50 and Inception-v3 fine-tuned on field weed photos with a heavy transform chain.
• Somepalli et al. (SAINT): row-attention transformer for tables, pre-trained with CutMix and mixup views and a contrastive loss.
• Li Jing et al. (dimensional collapse): shows that contrastive embeddings collapse into a low-dimensional subspace, and ties strong augmentation to it in linear models; proposes DirectCLR.
• Rakhlin et al. (breast histology): 50 colour and 20 crop views per slide, pooled over frozen ImageNet encoders, with a LightGBM classifier.
• Wang & Qi (CLSA): strong views trained to match the similarity profile of their weak twin over a key queue, added to contrastive pretraining.
• Li, Hou & Che (NLP augmentation survey): survey of text augmentation into paraphrasing, noising and sampling; no experiments of its own.
• dos Santos & Papa (CNN regularization survey): survey of CNN regularizers as input, internal and label methods; compiled results not reproduced.
• Yang et al. (image augmentation survey): survey of image augmentation with its own with and without runs on segmentation, classification and detection.
• Kong et al. (LaserMix++): LiDAR band swap between scans with a teacher, extended with camera correspondences and distillation.
• Safieh et al. (Jordanian speech): noisy-student loop for speech recognition, with Gaussian noise, time stretch and pitch shift at each generation.
• Song et al. (few-shot survey): survey that groups few-shot augmentation into hand-crafted and learned methods; no experiments of its own.
• Blanco-Gonzalez et al. (AI in drug discovery review): narrative review that names synthetic data as a remedy for scarce data; no method or experiment.
• Khan et al. (imbalance benchmark): nine resamplers paired with nine ensemble models on imbalanced binary and multi-class sets.
• Teye et al. (emotion detection): names data augmentation and oversampling without specifying them.
• Hao et al. (raindrop removal): ray-traced raindrops with exact masks rendered onto Cityscapes photos, and a three-stage detect, reconstruct and refine network.
• Howard & Gugger (fastai): layered deep learning library whose batch transforms run on the GPU; augmentation is one part of a training API, with no accuracy study.
• Zhu et al. (GCA): graph contrastive learning with edge and feature-dimension drops set by node centrality.
• Suresh et al. (AD-GCL): a GNN augmenter learns per-edge drop probabilities against a GIN encoder for graph contrastive learning, with a penalty on the drop ratio.
• Wang et al. (video saliency): synthetic video made by warping annotated still images with superpixel flow, used to train a fully convolutional saliency network.
• Vaze et al. (open-set recognition): maximum-logit scoring with a strong closed-set recipe that uses RandAugment and label smoothing.
• Lyu et al. (RTMDet): real-time one-stage detector family, with cached Mosaic and MixUp and a large-scale jittering stage.
• Ji et al. (OXE-AugE): cross-paints 16 Open X-Embodiment datasets onto up to nine robot arms by segmentation, inpainting and simulation replay.
• Yan et al. (InstruGen): route-and-instruction pairs from house-tour videos, with instructions written by a multimodal model and checked against the route.
• Yan et al. (OLiDM): object-first LiDAR diffusion, with generated foreground objects placed into a range-image scene model and a per-category loss.
• Wang & Wang (two-stream skeleton RNN): random 3D rotation, scaling and shear of skeleton sequences for a two-stream RNN action recogniser.
• Zhao et al. (self-consistency deepfake): stitched face forgeries with elastic masks, used as training data for a patch-consistency deepfake detector.
• Zhang et al. (non-Lambertian depth): random tone-mapping brightness curves and a gradient-domain loss for depth on glass and mirror surfaces.