Object detection

Overview

An object detector outputs, for each object it finds, a class label, a box and a confidence score. Training matches predicted boxes to annotated ones, and evaluation counts a match when the overlap (IoU) passes a threshold, typically 0.5 1802.09567 §5 1801.09454 §5.

The invariant every variant keeps: the output is a set of labelled boxes, and each annotated object should be explained by one of them. The variants differ in how candidate boxes are generated, how each object is matched to candidates during training, and how duplicates are removed. Dense detectors generate many candidates and suppress duplicates with non-maximum suppression (NMS). Set-prediction detectors learn a fixed set of query slots and match them to the ground truth during training 2211.12860 §1 2011.10881 §1.

The cards name SSD, YOLO, Faster R-CNN, RetinaNet, EfficientDet, ATSS, FCOS and DETR as published detectors, and in 3D they name BEVFormer, DETR3D and PointPillars. The wiki has no page for those models yet, so they are described here only through the cards that use them.

Variants

Two-stage region proposals (Faster R-CNN)

A region proposal network first proposes candidate boxes from the backbone features, then a second stage classifies each proposal and regresses its box. The variant in Learning Rich Features for Image Manipulation Detection is a Faster R-CNN with two ResNet-101 streams. The proposals come from the RGB stream only, because proposals from the noise stream performed worse 1805.04953 §3.1 1805.04953 tab. 1. Test-time NMS uses an IoU threshold of 0.2 1805.04953 §3.4. The detector outputs boxes, not masks, so it targets tampered regions, which need not be objects 1805.04953 §1 1805.04953 §3.4.

• The two streams are fused after RoI pooling, which gives a pixel F1 of 0.722 on NIST16, against 0.567 for the RGB stream alone and 0.625 for late fusion 1805.04953 tab. 3.
• Two full ResNet-101 trunks mean backbone compute roughly doubles compared with one stream. The paper does not report runtime 1805.04953 §3.2.

Card: RGB-N (Zhou et al.).

The same detector family is the base of three other cards, each changing a different part of training.

• Tiny-object label assignment. RFLA replaces the overlap test used to pick positive locations with a distance between Gaussian receptive fields and Gaussian object models, followed by a second top-1 pass 2208.08738 §3.1 2208.08738 §3.3. On AI-TOD, Faster R-CNN rises from 11.1 to 21.1 AP, and DetectoRS with RFLA reaches 24.8 AP 2208.08738 tab. 5. The setup is ResNet-50, 12 epochs, one RTX 3090 2208.08738 §4.2.
• Semi-supervised training. Unbiased Teacher trains a Faster R-CNN with FPN and ResNet-50 on pseudo-labels from an exponential-moving- average teacher, with a focal loss on the RoI classifier [arxiv:2102.09480 §3.2] 2102.09480 §3.3. With 1% of COCO labels it reports 20.75 mAP, against 13.97 for STAC 2102.09480 tab. 1.
• Distillation. DIST is a student-teacher loss for dense outputs. On COCO, a Faster R-CNN R50 student goes from 38.4 AP alone to 40.4 with DIST and 41.8 with DIST plus feature mimicking, against 39.7 with classic KD 2205.10536 tab. 6. The same table gives 41.7 for CWD, so DIST is not the best of the methods compared there.

Single-stage dense detectors (SSD, YOLO, RetinaNet)

A single network predicts classes and box offsets on a grid of anchors or points, with no second per-proposal stage. The cards in this group show the trade-off in three settings: a small detector for a fixed set of classes, a cascade of small detectors for a whole pipeline, and a multi-scale fusion change to SSD.

Road Damage Detection Using Deep Neural Networks with Images Captured Through a Smartphone trains SSD with MobileNet and SSD with Inception V2 at 300×300 input, on 9,053 dashboard-camera images with eight damage classes 1801.09454 §4 1801.09454 §3.4. MobileNet runs in 30.6 ms per image on a GPU. The same model takes 1.5 s on a Nexus 5X phone, measured with an older TensorFlow graph 1801.09454 tab. 3.

Real-time ALPR with YOLO (Laroca et al.) chains four YOLO-style detectors: vehicle, plate, characters, then letter or digit. The plate detector is Fast-YOLO 1802.09567 §4.1 [arxiv:1802.09567 tab. 1], and character segmentation uses CR-NET, a Fast-YOLO-inspired network 1802.09567 §4.2. Vehicle detection on the moving-camera dataset reached at most 97.33% recall with Fast-YOLO, so the authors switched that stage to YOLOv2 1802.09567 §5.2.1. The full pipeline runs at 47 FPS on a Titan Xp for the static-camera dataset 1802.09567 tab. 5 and 35 FPS on the moving-camera dataset 1802.09567 tab. 7.

FSSD keeps SSD's structure and fuses three VGG16 maps (conv4_3, fc7, conv7_2) into one feature map before the detection pyramid 1712.00960 §3.1 1712.00960 §4.2. The authors report 78.8 mAP on VOC2007 test at 65.8 FPS on a 1080Ti, and 82.7 with COCO pre-training 1712.00960 tab. 3. They say the fusion costs a small amount of speed, without a number in the card 1712.00960 §4.

• Per class, recall is uneven. In the road-damage baseline the rarest classes, D11 and D40, have low recall, which the authors link to few training examples 1801.09454 §5 1801.09454 tab. 2.
• A weak early stage caps the whole pipeline. The only vehicle missed on the moving-camera dataset was lost in the plate stage, where a false positive outscored the real plate 1802.09567 §5.2.2.
• Speed and accuracy are both reported on one colonoscopy benchmark 2011.07631 tab. III. YOLOv4 reaches AP 0.8513 at 48 FPS and RetinaNet with ResNet-101 reaches 0.8745 at 16.8 FPS. Faster R-CNN with ResNet-50 reaches 0.7866 at 8 FPS, and EfficientDet-D0 reaches 0.4756 at 35 FPS. The authors name YOLOv4 as the best detector for the balance of accuracy and speed 2011.07631 §VI.

YOLO-LITE is a seven-layer network with no batch normalisation, found by trial and error from Tiny-YOLOv2 for computers without a GPU 1811.05588 §3.1 1811.05588 tab. VII. On PASCAL VOC 2007 validation it reaches 33.57% mAP at 21 FPS on one laptop without a GPU, against 40.48% and 2.4 FPS for Tiny-YOLOv2 1811.05588 tab. III. Its COCO model reaches 12.26% mAP at the same speed 1811.05588 tab. V. Speed is measured on one device, and the authors call COCO-level accuracy the limit for real applications 1811.05588 §7.

microYOLO is a YOLO-style detector for a Cortex-M7 microcontroller with no accelerator. It takes 128 x 128 input, uses a small depthwise-separable backbone, and is then pruned with an L1 criterion and quantized to 8 bits 2408.15865 §2. On the OpenMV H7 R2 the fridge model uses 752 KB of Flash and 324 KB of RAM at 3.46 FPS 2408.15865 tab. 2. Author-reported mAP50 is 56.4% on the fridge task, 27.7% on humans and 12.3% on vehicles 2408.15865 §3.1 2408.15865 fig. 1. Shrinking the vehicle input to 88 x 88 cuts Flash to 133 KB and RAM to 152 KB, and raises speed to 8.40 FPS 2408.15865 tab. 3. The authors call the work in progress 2408.15865 §1.

Edge YOLO is a slimmed YOLOv4-style detector for a Jetson board in a vehicle, with a cloud server that retrains it and sends new weights back 2205.14942 §III-A. Its CSPDarknet53 backbone is cut to 17 layers from 88, the neck is an FPN without PANet, and there are three heads at 13, 26 and 52 cells 2205.14942 §III-B 2205.14942 tab. 1. On COCO2017 it reports 0.473 mAP@0.5 at 26.6 FPS on Xavier, against 0.657 at 2.1 FPS for YOLOv4 2205.14942 tab. 5. Under TensorRT FP16 the same table gives 0.325 at 49.5 FPS. The 26.6 FPS figure comes from a run whose precision the text does not state 2205.14942 tab. 5. The Soft-NMS threshold, σ and decay function are not stated, so the post-processing cannot be reproduced exactly 2205.14942 alg. 2.

DDOD keeps a one-stage FPN detector and separates three parts of its training: classification and regression choose their own positives with a cost-based assignment, the first convolution of each head branch is deformable, and each pyramid level is weighted by how many positives it has received 2107.02963 §4.1 2107.02963 §4.2 2107.02963 §4.3. On ATSS, the three parts together raise COCO val AP from 39.4 to 41.6, and each part alone adds 0.6 to 1.2 AP 2107.02963 tab. 4. The same recipe gives +2.0 to +2.4 AP on RetinaNet, FCOS and ATSS 2107.02963 tab. 3. The released code runs the assignment twice per iteration, and the paper gives no matched speed comparison 2107.02963 tab. 8.

YOLO-Z changes YOLOv5 for small objects, such as traffic cones seen from a racing car. Its strongest change replaces the lowest-resolution feature map with a higher-resolution one, which the authors call XS_ex 2112.11798 §3.3.3. At the S scale, small-object mAP@0.5 rises from 0.869 to 0.925 while inference goes from 8 to 8.9 ms on a GTX 1080 Ti [arxiv:2112.11798 tab. 2]. Feature-map changes gave the largest small-object gain at every scale 2112.11798 §4.3. The results come from one dataset, with single runs.

YOLO5Face treats face detection as generic object detection. It adds a five-point landmark head to YOLOv5, a stride-64 output for large faces, and ShuffleNetV2 backbones for the smallest models 2105.12931 §3.2. On WiderFace validation, the largest model reports 96.67, 95.08 and 86.55 (Easy, Medium, Hard), and the 1.7M-parameter YOLOv5n reports 93.61, 91.54 and 80.53 2105.12931 tab. 3. The paper gives FLOPs and parameters, not measured latency 2105.12931 §5.

Cards: Road Damage Detection, Automatic License Plate Recognition, FSSD, Real-Time Polyp Detection, Localization and Segmentation, YOLO-LITE, microYOLO, DDOD, YOLO-Z, YOLO5Face, Edge YOLO.

Box regression loss on auxiliary boxes (Inner-IoU)

The box regression loss sits on the box head of any dense or two-stage detector, so this change does not depend on the architecture. Inner-IoU replaces only the IoU term of a CIoU or SIoU loss. It scales the predicted and ground-truth boxes about their own centres, takes the IoU of the scaled boxes, and keeps the penalty terms on the full boxes 2311.02877 §3.2 2311.02877 eq. 15-27. The authors argue that a ratio below 1 sharpens the gradient for high-IoU samples and a ratio above 1 does the same for low-IoU samples 2311.02877 §3.1.

• On VOC with YOLOv7-tiny, Inner-CIoU at ratio 0.70 reaches 38.38 mAP50:95 against 37.64 for CIoU, and Inner-SIoU at 0.75 reaches 38.52 against 37.31 for SIoU 2311.02877 tab. 1. Every VOC row is above its baseline, by 0.61 to 1.21 points of mAP50:95 2311.02877 tab. 1.
• On AI-TOD with YOLOv5s the gains are mixed. At ratio 1.10 mAP50:95 falls by 0.17 while AP50 rises by 0.72; at ratios 1.13 and 1.15 mAP50:95 rises by 0.17 2311.02877 tab. 2.
• The paper gives no rule for choosing the ratio and no validation split, and each row is a single run 2311.02877 §4.2. Only CIoU and SIoU were tested, although GIoU, DIoU and EIoU are written into the method [arxiv:2311.02877 §4.2] 2311.02877 eq. 23-27.
• With a ratio below 1, two overlapping boxes can have disjoint auxiliary boxes. The IoU term then has zero gradient, a case the paper does not discuss 2311.02877 eq. 19 innerciou.py.

Card: Inner-IoU.

Real-time one-stage design study (RTMDet)

RTMDet is a family of one-stage detectors, from tiny to extra-large, built by testing design choices one at a time for real-time use 2212.07784 §1. Three choices carry most of the design. The basic block uses a 5 by 5 depthwise convolution, which the authors prefer over 3 by 3 and 7 by 7 for its speed-accuracy balance 2212.07784 tab. 5. The network has fewer blocks per stage and wider channels, and the neck gets a larger expansion ratio so that backbone and neck have similar capacity 2212.07784 §3.2. Training uses a dynamic label assignment in which the classification target is the IoU between prediction and ground truth, and the regression cost is the negative log of that IoU 2212.07784 §3.3 2212.07784 eq. 2 2212.07784 eq. 3.

• Reported on COCO val2017 with the authors' own latency setup: FP16, TensorRT, batch 1, one NVIDIA 3090, NMS excluded 2212.07784 §4.1. RTMDet-x reaches 52.8 AP at 3.10 ms. RTMDet-s reaches 44.6 AP at 1.22 ms, against 43.5 AP at 0.92 ms for YOLOv6-s, which the table lists with a letterboxed input 2212.07784 tab. 2. So RTMDet-s is more accurate and slower than that YOLOv6-s row, not faster.
• Mosaic and MixUp need several images per sample, which costs loading time. Keeping a cache of already-loaded images cuts that cost, with cached Mosaic at 24.0 ms per 100 images against 87.1 ms uncached 2212.07784 §3.3 2212.07784 tab. 7. A large cache with random eviction behaves like uncached augmentation, while a small cache with first-in-first-out eviction repeats images within a batch 2212.07784 §3.3.
• The same family extends to instance segmentation and to rotated boxes, each with small changes to the output part of the network 2212.07784 §3.4. The paper's rotated results are on DOTA 2212.07784 tab. 4.

NMS-free YOLO (YOLOv10)

YOLOv10 keeps the YOLO layout but removes NMS at inference. It trains two heads with the same structure: a one-to-many head with rich supervision, and a one-to-one head that makes the final predictions. A shared matching metric keeps the two consistent 2405.14458 §3.1. The authors then cut cost in several components, stage by stage [arxiv:2405.14458 §3.2]. On COCO val, the N to X models reach 38.5 to 54.4 AP, with end-to-end latency from 1.84 to 10.70 ms on a T4 with TensorRT FP16 [arxiv:2405.14458 tab. 1]. The paper's own remaining gap is 1.0 AP for N and 0.5 AP for S, against one-to-many training 2405.14458 §4.3 2405.14458 app. A.9.

YOLO family comparisons and reviews

Six cards compare or review YOLO versions. They do not share a protocol, so their rankings are not directly comparable.

• YOLO Evolution benchmarks 33 Ultralytics checkpoints, from YOLOv3u to YOLOv12, on three datasets: traffic signs, African wildlife and ships. All use one training setting, with one run per model [arxiv:2411.00201 §3.2] 2411.00201 tab. 3. Its traffic-sign table puts YOLO11m at mAP50 0.893 and mAP50-95 0.795 2411.00201 fig. 12. Averaged over scales and datasets, YOLO11 and YOLOv12 tie for first on accuracy, and YOLOv12 ranks last on speed 2411.00201 tab. 4.
• YOLO11 and YOLOv8 on orchard fruit trains YOLO11-seg and YOLOv8-seg in five sizes each, on 991 images of immature apples [arxiv:2410.19869 §3.2]. YOLO11m-seg has the best box mAP@50 overall, 0.876, and YOLOv8n-seg the lowest inference time, 3.3 ms 2410.19869 tab. 2 [arxiv:2410.19869 tab. 3]. Post-processing differs widely: 56.7 ms for YOLO11n-seg against 2.6 ms for YOLO11s-seg. The speed ranking therefore depends on which stage is measured.
• What is YOLOv8 is a descriptive review. It runs no experiments, and its YOLOv5 against YOLOv8 table gives no dataset or hardware 2408.15857 tab. 1. Use it for architecture, not for numbers.
• A review of YOLO architectures from YOLOv1 to YOLOv8 covers sixteen versions, with a summary table of reported COCO AP. The table gives 33.0 for YOLOv3 and 55.8 for YOLOv5, while the text gives 36.2 for YOLOv3-spp and 50.7 for YOLOv5 at 640 pixels 2304.00501 tab. 4 2304.00501 §6.3 2304.00501 §9. The paper does not run any model itself 2304.00501 §8.
• YOLO11 architectural overview describes C3k2 and C2PSA blocks and one model family from nano to xlarge. It runs no experiments. Its accuracy and latency figures are read from a third-party chart and given as approximations 2410.17725 §6 2410.17725 fig. 2.
• YOLO advances to its genesis reviews YOLOv1 to YOLOv12 and five variant detectors, walking the versions in reverse order. It runs no experiments, so every number is from a cited work 2406.19407 abstract. The paper states that no datasets were generated or analysed [arxiv:2406.19407 p. 68]. Its keywords name YOLOv13 to YOLOv20, which the body does not cover, so those versions are not reviewed here 2406.19407 abstract.

Set prediction with transformers (DETR family)

Each object is assigned to one query, and the model learns to output one box per query with no NMS at inference. The cards disagree on why DETR-style training is slow, and they propose different fixes.

TSP-FCOS and TSP-RCNN remove DETR's decoder and run a six-layer transformer encoder over about 700 features of interest, which come from FCOS or an RPN 2011.10881 §4. The authors name the cross-attention module as the main cause of slow convergence 2011.10881 §1 2011.10881 §3.2. With ResNet-50 at 36 epochs they report 43.1 and 43.8 COCO val AP, against 42.0 for DETR after 500 epochs 2011.10881 tab. 1.

H-DETR keeps one-to-one matching for the final model and adds a training-only group in which each ground truth appears K times, with K=6 for COCO detection. Masked self-attention keeps the two groups apart 2207.13080 §3. The authors note that Deformable-DETR selects fewer than 30 of its 300 queries as positives per image 2207.13080 §1. Deformable- DETR with ResNet-50 at 12 epochs goes from 47.0 to 48.7 AP [arxiv:2207.13080 tab. 1]. Inference cost is unchanged 2207.13080 §3.

Co-DETR adds auxiliary detection heads (ATSS and Faster R-CNN) on the encoder output, with one-to-many label assignment, and feeds their positive boxes back into the decoder as extra query groups 2211.12860 §3.2 2211.12860 §3.3. Its inference path is the plain set predictor.

• On Deformable-DETR with a ResNet-50 backbone at 12 epochs, COCO val AP goes from 37.1 without auxiliary heads to 42.3 with one ATSS head and 42.9 with two heads 2211.12860 tab. 2.
• The extra training cost is substantial. For Deformable-DETR++ with ResNet-50, GPU hours go from 70 to 86 with one head, 120 with two, and 280 with six 2211.12860 tab. 7.
• The strongest published Co-DINO results use NMS (soft-NMS in the released configs), so they are not NMS-free 2211.12860 §4.3.

OrientedFormer applies a DETR-style decoder to oriented boxes in remote-sensing images. It encodes each box as a Gaussian for the positional encoding, biases decoder self-attention with Gaussian Wasserstein distances, and samples cross-attention features at points rotated by the box angle 2409.19648 §3.3 2409.19648 §3.4 2409.19648 §3.5. With ResNet-50 it reports 67.28 AP50 on DIOR-R and 75.37 on DOTA-v1.0 single-scale, against 66.12 and 74.16 for ARS-DETR 2409.19648 tab. 3 2409.19648 tab. 4. It trains for 12 epochs against 36 for two end-to-end baselines, but uses about 2.8 times their FLOPs [arxiv:2409.19648 §4.8] 2409.19648 tab. 14.

PubTables-1M applies DETR without change to table images. One DETR finds the table on the page, with 15 queries, and a second predicts six structure classes, including rows, columns and spanning cells, with 125 queries 2110.00061 §4 2110.00061 app. Training Details. On the canonical test split, it reaches 0.8138 content accuracy against 0.1039 for Faster R-CNN, whose training details the card records as "default" only [arxiv:2110.00061 tab. 4]. The authors first merge oversegmented header cells, because the same table otherwise has two valid readings 2110.00061 §3. Scores on canonical and non-canonical test data are not comparable 2110.00061 §5.

Cards: TSP, H-DETR, DETRs with Collaborative Hybrid Assignments Training, OrientedFormer, PubTables-1M.

Attention-based heads and backbones

These cards keep a dense or two-stage detector and replace one of its parts with attention. Dynamic Head stacks blocks over a feature pyramid, viewed as levels, positions and channels. Each block applies three attentions in turn: one over scales, one over spatial positions through deformable sampling, and one per channel 2106.08322 §3.2. Plugged into ATSS with ResNet-50 at 1x, it moves COCO val AP from 39.4 to 42.6 [arxiv:2106.08322 tab. 3]. The gain also appears on Faster R-CNN, RetinaNet, FCOS and RepPoints, at 1.2 to 3.2 AP 2106.08322 §4.2 2106.08322 tab. 3. Six blocks add 21.5 GFLOPs to a 254.39 GFLOP baseline 2106.08322 tab. 2.

Swin Transformer is a backbone rather than a detector. It limits self-attention to local windows and shifts the window grid in every other block, so neighbouring windows exchange information 2103.14030 §3. Used with Cascade Mask R-CNN, Swin-T reaches 50.5 box AP on COCO val, against 46.3 for ResNet-50 2103.14030 tab. 2. The paper's shifted-window ablation gives 47.7 without shifting and 50.5 with it 2103.14030 tab. 4. Its speed row is on one V100 in images per second, 755 for shifted windows against 183 for a naive sliding window 2103.14030 tab. 5.

Conformer is a hybrid backbone: a CNN branch and a transformer branch run side by side and exchange features in every block through a Feature Coupling Unit 2105.03889 §3.2. In Faster R-CNN on COCO val, Conformer-S/32 reaches 43.1 box AP against 39.4 for ResNet-101 FPN, and in Mask R-CNN it reaches 39.7 mask AP against 36.1 2105.03889 §4.3 2105.03889 tab. 3. The ResNet-101 rows are taken from mmdetection, not re-run by the authors 2105.03889 tab. 3. Throughput is lower: 13.5 FPS against 15.9 for ResNet-101 in the Faster R-CNN setting, on a V100 at batch 1 2105.03889 app. C.

PVT v2 is a four-stage pyramid transformer backbone. Its attention pools keys and values to a fixed 7 by 7 grid, so the cost grows linearly with the number of tokens, while queries still come from every token 2106.13797 §3.2 2106.13797 eq. 2. With ATSS, the B2 model reaches 49.9 box AP on COCO val against 47.2 for Swin-T, and the linear B2-Li reaches 48.9 at 194 GFLOPs against 215 for Swin-T 2106.13797 tab. 4. With RetinaNet at 1x, B4 reaches 46.1 AP against 42.6 for PVT v1-Large 2106.13797 tab. 3. The comparison table does not state its schedule 2106.13797 tab. 4.

RFAConv replaces a k×k convolution with one whose weights differ at each window position. Each window gets its own softmax over its k×k slots, computed from pooled features, so the weights are not shared between overlapping windows 2304.03198 §3.3. In YOLO models the authors report COCO2017 AP rising from 36.4 to 37.7 for YOLOv8n and from 27.5 to 29.0 for YOLOv5n 2304.03198 tab. 7. Per-image validation time for YOLOv8n goes from 4.2 to 4.5 ms 2304.03198 tab. 7. The expanded tensor is k² times the input, a memory cost the authors state 2304.03198 §4.6.

Cards: Dynamic Head, Swin Transformer, Conformer, PVT v2, RFAConv.

Language-model decoders (VisionLLM)

VisionLLM writes detection as text for a language model to generate. A language-guided tokenizer turns the image into 100 learned tokens, shaped by the instruction. An Alpaca-7B decoder, adapted with LoRA, then writes class and coordinate tokens in the format the instruction names 2305.11175 §3.3 2305.11175 §3.4 2305.11175 §4.1. A box is two points, the top-left and bottom-right corners, and the category set is given in the instruction 2305.11175 §3.2 2305.11175 fig. 9.

• With a ResNet-50 backbone the model reaches 44.6 box AP on COCO, 1.4 above Pix2Seq at the same backbone. With InternImage-H it reaches 60.2 2305.11175 tab. 1. The ResNet-50 model trained separately for detection reaches 44.8 2305.11175 tab. 1.
• Masks are written as boundary points, and mask AP rises with their number: 18.5 with 8 points against 25.1 with 24 2305.11175 tab. 2. Box detection uses two points, so this trade-off does not apply to boxes.
• The authors argue that parallel decoding is more efficient than token-by-token output, but report no latency or throughput numbers 2305.11175 §4.2. The card records that the linked repository has no implementation of this method at the pinned commit, so the numbers cannot be checked against code card.

Oriented boxes in aerial images

ReDet makes an oriented-box detector rotation-equivariant instead of relying on rotated copies of the training data. Its backbone is a ResNet-50 with FPN, equivariant to eight discrete rotations, and RiRoI Align pools rotated proposals after aligning them in space and orientation [arxiv:2103.07733 §5.4] 2103.07733 tab. 1. On DOTA-v1.0 with multi-scale training and rotation augmentation it reports 80.10 mAP, against 79.42 for the best other row in its table 2103.07733 tab. 6. Its backbone is 12 Mb against 103 Mb for the baseline 2103.07733 tab. 1. The invariance is exact only at the grid angles, because RiRoI Align interpolates between neighbouring orientations 2103.07733 §4.1 2103.07733 §5.3. For a DETR-style alternative on the same kind of data, see OrientedFormer above.

Card: ReDet.

3D detection from point clouds, cameras and LiDAR-camera fusion

Several cards detect objects in 3D. They differ on what the detector sees, whether it is shared between vehicles, and how many cameras it uses.

Group-Free 3D detects indoor objects from point clouds without a hand-made grouping step. Each candidate attends to all points through stacked decoder layers, and the boxes from all layers are pooled and filtered with NMS at IoU 0.25 2104.00678 §3.2. The authors report 67.3 mAP@0.25 and 48.9 mAP@0.5 on ScanNet V2 for the best run, and 66.3 and 48.5 averaged over 25 trials 2104.00678 tab. 1. A 2D DETR-style setup on the same data reaches 39.6 and 21.4 after 400 epochs 2104.00678 tab. 10.

BADet is a LiDAR detector that refines its proposals jointly rather than one at a time. Proposals within a cut-off distance form a graph, and message passing between them updates each proposal's features. On KITTI test, Car, its moderate AP3D is 81.61, level with Voxel R-CNN at 81.62 2104.10330 tab. 1. Its average runtime is 140 ms on a GTX 1080 Ti 2104.10330 §4.5, which the authors give as 7.1 FPS against 1.5 for Point-GNN 2104.10330 tab. 9. With the graph refiner, KITTI val moderate AP3D rises from 79.94 to 85.77 over the same model without it [arxiv:2104.10330 tab. 5].

LaserNet runs one fully convolutional network on the LiDAR range image, with one row per laser, rather than on a bird's-eye-view grid 1903.08701 §3.1. Each point predicts a mixture of box distributions. Mean-shift clustering merges the predictions, and an adaptive NMS prunes them using the predicted spread 1903.08701 §3.4 1903.08701 §3.6. On the ATG4D validation set, vehicle AP at IoU 0.7 over 0 to 70 m is 85.34, against 80.99 for PIXOR 1903.08701 tab. 1. On KITTI the table gives 73.77 moderate, against 77.05 for PIXOR, and the authors say their method is worse than current BEV detectors there 1903.08701 tab. 4 1903.08701 §4.4. Total runtime is 30 ms on a 1080 Ti, with pre- and post-processing included 1903.08701 tab. 3 1903.08701 §4.3. The paper does not say how to obtain ATG4D, so its headline result may not be reproducible from the paper alone 1903.08701 §4.1.

AVOD fuses a LiDAR bird's-eye-view map with an RGB image in full-resolution feature pyramids. A region proposal network proposes 3D boxes from both views, and a second stage refines each box and predicts its orientation 1712.02294 §3.3 1712.02294 §3.4. On KITTI test, Car, the pyramid version reports 81.94, 71.88 and 66.38 AP3D for easy, moderate and hard, at about 0.1 s per frame on a TITAN Xp 1712.02294 tab. 2 1712.02294 §4. The feature-pyramid variant has the highest car scores, with the base network listed beside it 1712.02294 tab. 2. All numbers are author-reported.

BEVFusion gives the camera its own stream: image features are lifted into a bird's-eye-view grid and fused with a LiDAR BEV feature, so the camera path does not need LiDAR points 2205.13790 §3 [arxiv:2205.13790 §3.3]. On nuScenes validation with a TransFusion-L LiDAR stream, fusion gives 67.9 mAP against 64.9 for LiDAR alone 2205.13790 tab. 1. Under simulated LiDAR loss, the abstract reports margins of 15.7 to 28.9 mAP points over earlier methods 2205.13790 §4.4.1.

OPV2V is a benchmark for cars that share features over a radio link. Nearby cars share LiDAR features, so a car sees objects hidden behind other vehicles 2109.07644 §3.1. The benchmark compares four LiDAR detectors under no fusion, late fusion, early fusion and intermediate fusion 2109.07644 §5.1 2109.07644 tab. 4.

• For VoxelNet on the default towns, AP at IoU 0.7 rises from 0.526 with no fusion to 0.864 with intermediate fusion 2109.07644 tab. 4.
• Compressing the shared features to 4096× costs about 3% AP and still beats early and late fusion. The authors estimate about 5 ms to send one message at 27 Mbps and 300 m 2109.07644 §5.6.
• All results are simulated. The paper does not give a license for the dataset, and the released code is non-commercial 2109.07644 §1.

V2V4Real is a real-world counterpart to simulated car-to-car data. Two instrumented cars drove together in Columbus, Ohio, and each car's LiDAR was labelled with 3D boxes 2303.07601 §3.1 2303.07601 §3.2. Its detection benchmark uses LiDAR only. There, CoBEVT reaches 66.5 AP@0.5 in the sync setting, against 39.8 for no fusion 2303.07601 tab. 3. Adapting a model trained on simulated OPV2V data lifts CoBEVT from 32.6 to 40.2 AP@0.5 2303.07601 tab. 5. The dataset is not openly licensed, so check its card before reuse.

Interpretable motion planner feeds LiDAR sweeps and an HD map through one network that also predicts vehicle detections and motion 2101.06679 §3.1. Detection is one head of a planner, so its own accuracy is not the headline; the paper reports a lower future collision rate than the compared planners on a private dataset 2101.06679 §4.

VoTr swaps the sparse-convolution backbone of a voxel detector for Transformer blocks. Each occupied voxel attends only to other occupied voxels, at most 48 per query, chosen from a near range and a dilated range with larger steps 2109.02497 §3.2 2109.02497 §3.3 2109.02497 §4.1. A GPU hash table finds those neighbours 2109.02497 §3.4. Its single-stage form, VoTr-SSD, reports 78.25 AP3D moderate on KITTI test, against 75.96 for SECOND 2109.02497 tab. 2. The authors' runtime is 14.65 Hz against 20.73 Hz for SECOND [arxiv:2109.02497 tab. 8]. The code repository has no license file, so check the card before reusing any of it.

Two cards detect objects from several cameras or from cameras and LiDAR together. PolarFormer predicts objects on a polar grid of distance and azimuth, fed by six surround cameras 2206.15398 §3.1 [arxiv:2206.15398 §3.4]. On nuScenes validation with a ResNet-101 backbone, it reports 39.6 mAP and 45.8 NDS, against 37.5 and 44.8 for BEVFormer-S 2206.15398 tab. 3. Its 50.0 mAP setting uses a VoVNet backbone and extra pretraining data [arxiv:2206.15398 §4]. The paper reports no runtime 2206.15398 §4.

DeepInteraction keeps a LiDAR map and the camera features as separate representations, and lets them exchange information through projections. A two-layer encoder does the exchange, and a decoder reads the image and LiDAR maps in turn 2208.11112 §3.1 2208.11112 §3.2. The base model, with a ResNet-50 image backbone, reports 70.8 mAP and 73.4 NDS on nuScenes test. Its LiDAR-only baseline, TransFusion-L, reports 65.5 mAP on the same split 2208.11112 tab. 1. DeepInteraction runs at 2.6 FPS on a V100, against 3.8 for TransFusion 2208.11112 tab. 2. The method depends on sensor calibration 2208.11112 app. A.3.

CenterFusion adds radar to a camera-only, center-point 3D detector. Radar points are expanded to pillars, the closest one inside each object's frustum is kept, and its depth and velocity are painted onto image-shaped maps for second-stage heads 2011.04841 §4.2 2011.04841 §4.3. On nuScenes validation, both models with DLA, NDS rises from 0.328 to 0.453, while mAP rises only from 0.306 to 0.332 2011.04841 tab. 1. On the test split it reports NDS 0.449 against 0.400 for the Hourglass CenterNet, but mAP 0.326 against 0.338 2011.04841 tab. 1. The authors' ablation attributes most of the velocity error reduction to the frustum association 2011.04841 tab. 3. The paper and the released code differ in several parameters, such as the test-time region enlargement and the pillar size, so read the card before reproducing numbers 2011.04841 §5.

Card: CenterFusion.

Two more cards detect from camera images alone, with no LiDAR in the loop, and both lift image features into a bird's-eye-view layout. BEVFormer turns six surround-camera images into a BEV grid of 200 by 200 learnable queries with a six-layer encoder 2203.17270 §3.1. Spatial cross-attention reads image features where each grid cell's 3D points project, and temporal self-attention reads the previous BEV map 2203.17270 §3.3 2203.17270 §3.4. On nuScenes val with ResNet-101 it reports 51.7 NDS against 44.8 for BEVFormer-S 2203.17270 tab. 2. Its 56.9 NDS on test uses a VoVNet-99 backbone pre-trained on extra depth data 2203.17270 tab. 1. DETR3D uses no depth network and no NMS. Its 900 learned queries each hold a 3D reference point, which is projected into every camera to sample features 2110.06922 §3.3. It reports 0.374 NDS and 0.303 mAP on nuScenes val with ResNet-101 2110.06922 tab. 1. Its mATE is worse than FCOS3D's, and the authors suspect that FCOS3D's direct depth prediction supervises translation more strongly 2110.06922 §4.2. In a pseudo-LiDAR baseline the authors built, NDS is 0.160 2110.06922 tab. 4.

CaDDN detects cars from one camera. For every pixel it predicts a distribution over depth bins rather than a single depth value. Image features are weighted by those distributions, lifted into a frustum, sampled into a voxel grid, collapsed to bird's-eye view, and passed to a PointPillars style head, all trained end to end 2103.01100 §3.1 [arxiv:2103.01100 §3.2]. LiDAR gives the depth supervision during training only [arxiv:2103.01100 §3.4]. On KITTI test, Car, moderate AP R40, it reports 13.41 against 9.71 for M3D-RPN 2103.01100 tab. 1. On KITTI val, adding the depth-distribution supervision raises moderate Car AP from 6.43 to 14.03 2103.01100 tab. 3, and training the depth head jointly with the detector gives 13.71, against 12.26 when the depth is estimated separately 2103.01100 tab. 4. The released code implements the focal weighting differently from the wording in the paper, so read the card before porting it.

Pseudo-LiDAR++ estimates depth from a stereo pair with a network trained on depth rather than disparity. A few LiDAR points then correct the depth map, and their depths spread to nearby points through a graph 1906.06310 §3 1906.06310 §4. The paper argues that a disparity loss spends effort on near objects, because a fixed disparity error costs far more depth at long range 1906.06310 §3 1906.06310 eq. 5. On KITTI validation, car, moderate, IoU 0.7, PointRCNN, AP_BEV rises from 56.0 with the pseudo-LiDAR baseline to 64.0 with the new stereo network, and to 76.9 with four simulated LiDAR beams added 1906.06310 tab. 1. On the test split the four-beam setting reaches 73.5, against 85.7 for PointRCNN on 64-beam LiDAR 1906.06310 tab. 2. The four beams are thinned from 64-beam scans, and the depth ground truth for training still comes from 64-beam LiDAR, so the results do not show what a real four-beam sensor would give 1906.06310 §5.1.

Card: Pseudo-LiDAR++.

MonoDIS detects cars from one calibrated RGB image with a ResNet34 and FPN two-stage design 1905.12365 §4.1 [arxiv:1905.12365 §4.2]. Its main change splits the 3D box loss into one copy per parameter group. In each copy only that group comes from the prediction, and the rest is set to the ground truth 1905.12365 §5.1. A second head predicts a 3D confidence from the box error, without extra labels 1905.12365 §4.3. On KITTI test, car, moderate difficulty, 40-point AP reaches 7.94 with the larger training split and 4.89 with the standard split, against 5.74 for MonoGRNet on the same metric 1905.12365 tab. 3. The paper gives no official code [arxiv:1905.12365 §7.4].

TopoLogic predicts which lanes connect at an intersection, given lanes already detected as ordered point lists. It scores each lane pair twice: with a learnable function of the gap between one lane's end and the next lane's start 2405.14747 §3.3, and with a sigmoid of the similarity of their queries 2405.14747 §3.4. The two scores are fused with learnable weights and passed to a graph layer in the decoder 2405.14747 §3.5 2405.14747 §3.2. On OpenLane-V2 subset_A with a ResNet-50 backbone, it reports TOP_ll 23.9 and OLS 44.1 (v2.1.0), against 10.9 and 39.8 for TopoNet 2405.14747 tab. 1. The geometric term can also be added after training, which raises TopoNet's TOP_ll from 10.9 to 22.3 2405.14747 tab. 5. The released post-processing path uses fixed values in place of the learned parameters, so it does not reproduce the learned mapping [code: projects/topologic/models/dense_heads/topologic_head.py#L444-L452]. The authors say the work is for research only 2405.14747 §5.

A survey of 3D detection for driving groups methods by input: image, point cloud and fusion. It compares published KITTI, nuScenes and Waymo results 2106.10823 tab. 6 2106.10823 tab. 8, and its own case study runs fifteen detectors on KITTI validation. The case study finds that 3D location error contributes the most error 2106.10823 tab. 10. Its published tables are not re-measured, so they carry the same caveats as their source papers.

Cards: PolarFormer, DeepInteraction, Group-Free 3D, BEVFusion, AVOD, BADet, BEVFormer, DETR3D, 3D detection survey, OPV2V, End-to-end Interpretable Neural Motion Planner, VoTr, CaDDN, TopoLogic, MonoDIS, Pseudo-LiDAR++, CenterFusion.

Unknown objects and open-world detection

A closed-set detector labels an unseen object as one of its known classes, often with high confidence. Towards Open World Object Detection (ORE) treats unknown objects as something to flag now and learn later. The detector marks them as unknown, a human names new classes, and the model is updated without retraining from scratch 2103.02603 §3. ORE is a Faster R-CNN with three additions 2103.02603 §4:

• a contrastive clustering loss on RoI features, which pulls each class towards its own prototype 2103.02603 §4.1;
• pseudo-labels for unknowns, taken from high-objectness background boxes that do not overlap a ground-truth object 2103.02603 §4.2;
• an energy score, with two fitted Weibull distributions that decide whether a detection is unknown 2103.02603 §4.3.

Incremental steps replay a balanced set of old examples, with 50 per class chosen on validation 2103.02603 §4.4 2103.02603 §5.2.

• On the VOC then COCO task split, ORE lowers Wilderness Impact at task 1 from 0.0699 to 0.0219 against Faster R-CNN 2103.02603 tab. 2.
• At task 2, mAP over both known sets is 38.98 for ORE, against 37.47 for Faster R-CNN with finetuning and 14.91 without it 2103.02603 tab. 2. The no-replay baseline forgets most old classes, so the finetuning row is the fairer comparison.
• On the VOC incremental benchmark, ORE has the highest mAP in all three splits, 0.3 to 1.4 points above the next-best method 2103.02603 tab. 3.

Pseudo-labels come from background boxes, so a real but unlabelled object can be treated as unknown 2103.02603 §4.2. The released code differs from the paper in several hyperparameters, including the clustering margin and the energy temperature; the card lists them card.

Few-shot fine-tuning (FSCE)

FSCE keeps the ResNet backbone frozen and fine-tunes the FPN, the RPN and the RoI box head on K labelled shots of each novel class, plus base-class samples 2103.05950 §3 2103.05950 fig. 3. It adds a small head that encodes each sampled RoI, trained with a supervised contrastive loss. RoIs of one class are pulled together and the others pushed apart, and the term enters the total loss with weight 0.5 2103.05950 §3.2 2103.05950 §3.3 2103.05950 eq. 5. The authors argue that most few-shot errors are wrong class labels on boxes that are already well placed 2103.05950 §1.

• Without the contrastive term, RPN and RoI refinement alone reach 49.7 novel nAP50 at 3 shots on VOC split 1. The full method reaches 51.4 at 3 shots and 63.4 at 10 shots 2103.05950 tab. 5.
• On COCO, the 30-shot novel AP is 16.4, against 13.7 for TFA w/ cos 2103.05950 tab. 3.
• The released configs use other λ and IoU thresholds from the paper's text, and not every paper cell has a released config [code: configs/PASCAL_VOC/split1/3shot_CL_IoU.yml#L29].

Card: FSCE.

Long-tailed classes

When most training boxes belong to a few common classes, a sigmoid classifier sees far more negative than positive gradient for each rare class. EQL v2 reweights each class's positive and negative gradients from the ratio of their accumulated magnitudes, in one training stage 2012.08548 §3.2 2012.08548 eq. 5. On LVIS v1.0 with Mask R-CNN and ResNet-50 at 1x, it reports 23.7 AP and 14.9 AP on rare classes. The earlier EQL reports 18.6 and 2.1, softmax 16.1 and 0.0 2012.08548 tab. 1. The two-stage BAGS method reports 23.1 AP 2012.08548 tab. 1. On Open Images, the authors report 52.6 AP against 43.1 for Faster R-CNN, with the LVIS settings 2012.08548 tab. 10. The EQL row also differs by table: 18.6 AP in tab. 1 and 21.6 in tab. 5, which uses a 2x schedule and the authors' re-implementation 2012.08548 tab. 5.

Pixel masks for saliency and camouflage

These cards carry the detection tag, but they predict a mask per salient or camouflaged object rather than one box per object. Their scores are segmentation scores, so do not put them in a table with box AP. The salient object detection concept page covers them in more depth.

Video salient object detection runs a static network on each frame and a dynamic network on frame pairs plus the static map. No optical flow is computed at test time 1702.00871 §3.1 1702.00871 §3.3. Its training videos are synthetic, made by warping still images 1702.00871 §4. Training on image-derived video alone raises FBMS MAE from 7.65% to 9.27% 1702.00871 tab. 2, and the authors report about 2 fps for the full model 1702.00871 §5.4.

DVA predicts eye fixations with a VGG16 encoder and three deconvolution decoders, on conv3-3, conv4-3 and conv5-3, each with its own loss. A learned fusion layer merges the three maps 1705.02544 §3.3.1 1705.02544 §3.2. It trains on SALICON only and reports about 0.1 s per image on one GPU 1705.02544 §3.3.3. It leads on most metrics for four of the five eye-tracking sets, but not on MIT300, where DeeFix and SALICON score higher 1705.02544 tab. 4.

EGNet adds a salient-edge path built on the conv2-2 features and the top-level location. Its edge features are added to each object scale before a second prediction, and the four maps are fused [arxiv:1908.08297 §3.2.2] 1908.08297 §3.3. Max F-measure on DUTS-TE is 0.893 with ResNet and 0.880 with VGG 1908.08297 tab. 2. The paper does not name its optimiser 1908.08297 §4.1.

D3Net runs three streams: RGB only, RGB with depth, and depth only. At test time it keeps the RGB-plus-depth map only when it agrees with the depth-only map, within a mean absolute difference of 0.15, and otherwise uses the RGB map 1907.06781 §4.2 1907.06781 eq. 3. On NJU2K test its S-measure is 0.900, against 0.879 for CPFP 1907.06781 tab. 4. The 0.15 threshold was chosen on data the paper does not name [arxiv:1907.06781 §4.3].

SINet-V2 finds camouflaged objects in two stages. A neighbour connection decoder first gives a coarse map, and three group-reversal attention blocks then refine it 2102.10274 §4.2 2102.10274 §4.3. On the COD10K test set its S-measure is 0.815, on CAMO 0.820 and on CHAMELEON 0.888 2102.10274 tab. 2. Its dataset, COD10K, has 10,000 images 2102.10274 §3.1.

Anabranch Network multiplies one scalar into the map of any saliency network. The scalar is the probability that camouflage is present in the image, so an image without camouflage is pushed towards an empty map 2105.09451 §4.2. On CAMO-COCO, ANet-SRM reaches F-beta 0.826 and IoU 0.732, against 0.699 and 0.535 for the pre-trained SRM 2105.09451 tab. 3.

COD-Rank predicts a fixation map, a camouflage mask and a difficulty rank in one network. The ranks come from eye-tracker detection delays 2103.04011 §3.1 2103.04011 §3.2 2103.04011 §3.3. Its joint model trained on COD10K and CAMO reaches an S-measure of 0.793 on COD10K, against 0.776 for SINet 2103.04011 tab. 4. The released code has no ranking branch, so the rank results cannot be checked from it [code: joint_cod_col/train.py].

Cards: Video salient object detection, DVA, EGNet, D3Net, SINet-V2, COD-Rank, Anabranch Network.

Surveys

A Survey of Modern Deep Learning based Object Detection Models covers datasets and metrics, backbones, two-stage and single-stage detectors, and lightweight networks. Its detector table collects author-reported COCO or VOC AP and FPS for 17 detectors, without re-running any of them 2104.11892 §1 2104.11892 tab. 3. Its summary is that two-stage detectors are generally more accurate and single-stage detectors are faster 2104.11892 §9. Its headline rows are DetectoRS at 53.3 AP on COCO with a 1333 input, at about 4 FPS, and YOLOv4 at 43.0 AP with a 512 input, at about 31 FPS 2104.11892 tab. 3. The rows differ in hardware, input size and training data, so the survey calls them a guide rather than a ranking 2104.11892 §7. Its FPS column mixes devices, so measure speed on your own hardware.

Card: A Survey of Modern Deep Learning based Object Detection Models.

Image Data Augmentation for Deep Learning: A Survey groups augmentation into manipulation, erasing, mixing, automated policy search, feature-space and GAN methods 2204.08610 §2 2204.08610 §3. Its own experiments use only the first three groups, on segmentation, classification and detection 2204.08610 §4. For detection on COCO2017, Faster R-CNN rises from 36.40 to 36.80 mAP with augmentation, and CenterNet falls from 41.42 to 41.15 2204.08610 tab. 4. No training settings or run-to-run variance are reported 2204.08610 §4.3.

Moving Forward is a survey of autonomous-driving sensors, software and hardware. Its object detection table lists Faster R-CNN, YOLO, SSD, VoxelNet, PointPillar and other 2D and 3D detectors, with mAP and FPS taken from the original works and not re-measured by the authors 2411.10291 tab. II. The table mixes VOC2007 2D results with KITTI 3D results, and it does not say what hardware the FPS came from, so it is a reference, not a ranking 2411.10291 tab. II. The authors' own measurements cover end-to-end driving models, not detection alone 2411.10291 §V-A.

A Review on Deep Learning in UAV Remote Sensing screened 232 articles and proceedings from 2016 to 2021 and tallies them by network, task, sensor and setting. Object detection is 53.9% of the coded studies and segmentation 40.7%. The authors' tallies are not independent measurements 2101.10861 §3. Detection studies mostly use RGB imagery 2101.10861 §3.

A Survey of Deep Learning Techniques for Weed Detection from Images reviews 70 deep-learning papers from 2016 to August 2020 on finding weeds in crop images. Its detector list covers Faster R-CNN, SSD, Mask R-CNN and YOLO-v3 and tiny YOLO-v3 2103.01415 §10.2. Most of the 19 public datasets are specific to one site or crop 2103.01415 §6.3. The reported scores come from different crops, metrics and datasets, and the survey says they are hard to compare 2103.01415 §11.

3D Object Detection for Autonomous Driving: A Comprehensive Survey organises 3D detection by sensor: LiDAR, camera and multi-modal 2206.09474 §3 2206.09474 §4 2206.09474 §5. It concludes that LiDAR and fusion methods are the choice when speed and accuracy matter most, and camera methods when cost matters most 2206.09474 §10.1.6. Its tables copy scores from original papers that use different KITTI, nuScenes and Waymo splits, so they are not one ranking 2206.09474 tab. 16 2206.09474 tab. 17.

Benchmarks and datasets

Some cards are datasets or competitions rather than methods. Their baselines give a starting point and a metric, not a ranking of methods.

xView is an overhead dataset with over one million boxes in 60 classes. Its SSD baseline reaches total mAP 0.1456 on 300-pixel chips and 0.2590 with chips of three sizes 1802.07856 tab. 1. Augmentation on top of the three sizes gives 0.1549, so it does not help in this setup 1802.07856 §4.

The UAV maritime benchmark provides detection and tracking labels for swimmers, life jackets and boats, with per-frame altitude and camera angle. The best detector in its baselines is Faster R-CNN with ResNeXt-101-FPN at AP 30.4, and EfficientDet-D0 reaches 20.8 2105.01922 tab. 4. The test labels are withheld and scored on a server 2105.01922 §3.3.

The ICDAR 2019 receipt competition includes a text-line localisation task. The top entry reaches Hmean 98.59% 2103.10213 §III-A2 2103.10213 fig. 3. This is detection of text rather than objects, so its numbers are not comparable with the box-detection results above.

ARKitScenes records about 5,000 RGB-D captures of 1,661 indoor scenes with a tablet's LiDAR sensor, with oriented 3D boxes for 17 furniture classes 2111.08897 §3. Three indoor detectors trained on it reach whole-scene mAP of 0.358 (VoteNet), 0.383 (H3DNet) and 0.419 (MLCVNet) at IoU 0.25 2111.08897 tab. 2. The per-frame setting changes the order, with VoteNet first 2111.08897 tab. 3. All numbers are author-reported.

RDD2022 is a road-damage dataset of 47,420 photos from six countries, with 55,007 labelled boxes in the training split for four classes: longitudinal and transverse cracks, alligator cracks and potholes 2209.08538 tab. 2 2209.08538 fig. 5. Test labels are withheld and scored through a leaderboard 2209.08538 p. 11. The paper reports dataset counts only, with no detector results 2209.08538 p. 1.

Synscapes is a synthetic street-scene set of 25,000 path-traced images, with instance maps and per-object 2D and 3D boxes 1810.08705 §3. Faster R-CNN trained on it alone reaches 0.488 mAP on its own validation split 1810.08705 tab. 6. Adding it to KITTI training raises KITTI-validation mAP from 0.456 to 0.519 1810.08705 tab. 6. The paper does not state the dataset's licence or download location, so check its terms before reuse 1810.08705 §3.

Cards: xView, Synscapes, UAV maritime benchmark, ICDAR 2019 receipts, ARKitScenes, RDD2022.

Detection as a component of other tasks

Some cards use detection as one step in a larger system. Their headline results measure the larger task, not the detector.

POI tracks people with a Faster R-CNN detector and a 128-dimensional appearance embedding trained with softmax and triplet losses 1610.06136 §1 1610.06136 §2. The authors show that detection errors carry into tracking. Adding skip pooling and multi-region features cut false positives plus false negatives on MOT16 train from 52,727 to 46,587 1610.06136 tab. 1. On the MOT16 challenge, the online tracker reaches MOTA 66.1 and the offline one 68.2, both author-reported 1610.06136 tab. 4.

YOLOP runs one shared encoder with three heads, for vehicles, drivable area and lane lines 2108.11250 §3. Only vehicles are scored for detection 2108.11250 §4.2.1. Its jointly trained model reaches 76.5 vehicle mAP50 at 41 fps on one TITAN XP 2108.11250 tab. 1. One multi-task pass takes 24.4 ms, against about 45 ms for the three single-task models summed from the same table 2108.11250 tab. 5.

YOLO-Pose adds 51 keypoint values per anchor to a YOLOv5-style head. Each person's box and 17 keypoints then come from one pass, with no grouping step 2204.06806 §3 2204.06806 eq. 1. The keypoint loss is one minus the object keypoint similarity 2204.06806 §3.4 2204.06806 eq. 3. The largest model reaches 90.2 AP50 on val2017, but its AP there is 69.4, below 69.9 for HigherHRNet-W48 at 640 2204.06806 tab. 1.

MARTENS places an oriented-box YOLOv8 detector inside a simulated pick-and-place cell. A genetic search over box poses and lighting looks for scenes where the arm fails, and the detector is then fine-tuned on the failing images 2410.19277 §2.2 2410.19277 §2.5. Author-reported: the search finds more failures than random search in both use cases [arxiv:2410.19277 tab. 2] 2410.19277 tab. 3, and fine-tuning repairs more than 99% of them 2410.19277 tab. 6. A detector that starts from the repaired simulation model reaches 0.948 and 0.892 mAP after training on about 100 real images per use case. A real-only detector reaches 0.806 and 0.755 2410.19277 tab. 8. Each simulated test takes 280 to 310 s on one GPU, so the search is costly 2410.19277 §3.1.

The agricultural deep-learning survey tabulates 40 studies. Weed identification and fruit counting are among its application areas, with five and four studies respectively 1807.11809 §4.1. Most of its reported numbers are classification accuracy, not detection, and setups differ across studies, so they are not directly comparable 1807.11809 §4.9.

Florence-2 handles detection, grounding and captioning in one prompt-driven model, with boxes written as location tokens 2311.06242 §3. Its detection results are reported as one task among many 2311.06242 tab. 4.

The scene graph survey covers the step after detection, where objects are linked by relations. Its advice is to report mean recall per predicate, because plain recall is dominated by common relations 2104.01111 §6.2.

The autonomous racing survey lists high-speed object detection and localisation without HD maps as open problems 2202.07008 §2.1. The remote sensing transformer survey reproduces detection numbers from the original papers, such as DOTA mAP up to 80.14 2209.01206 §5. Neither card re-runs its numbers.

Visual-RFT fine-tunes a vision-language model with GRPO, using a reward that combines the IoU of each predicted box with its matching ground-truth box, plus a confidence term 2503.01785 §3.2.1 [arxiv:2503.01785 eq. 5]. On 8 COCO categories with two images each, the 2B model reaches 41.5 mAP, against 21.0 for supervised fine-tuning and 19.6 for the base model 2503.01785 tab. 3. The paper does not state the IoU threshold used in the reward 2503.01785 §3.2.1.

The anomaly detection survey for driving lists lidar object-level anomaly detection as only just gaining momentum 2204.07974 §3. It runs no experiments of its own.

The social-distancing monitor runs a YOLO v3 person detector with DeepSORT tracking, then groups people who stand close together 2005.01385 §IV-A. On the authors' 800-image person set, YOLO v3 reaches 0.846 mAP at 23 FPS, against 0.969 for Faster R-CNN at 3 FPS and 0.691 for SSD at 10 FPS 2005.01385 tab. III. The authors call YOLO v3 balanced across mAP, training time and speed 2005.01385 §V. The paper does not state the IoU threshold behind its mAP, and it never checks the grouping or the violation index against ground truth 2005.01385 §VI.

SportsMOT is a tracking dataset for players in basketball, volleyball and football. Its trackers take per-frame boxes from a YOLOX detector 2304.05170 §5.1. The authors report that the Kalman motion model lowers HOTA on this data, 64.1 with the filter against 71.5 without it [arxiv:2304.05170 tab. 6].

GenEval uses a detector as a checker for text-to-image output. A Mask2Former with a Swin-S backbone, trained on the 80 COCO classes, finds the objects in each generated image, and a confidence threshold of 0.3 decides which detections count, or 0.9 for counting 2310.11513 §3.2 2310.11513 app. C.3. The authors state that COCO-trained detectors cannot verify fine details or artistic renderings 2310.11513 §6. On the paper's six-task overall score, IF-XL reaches 0.61, with position at 0.13 and attribute binding at 0.35 2310.11513 tab. 2.

Cards: POI, YOLOP, YOLO-Pose, Social-distancing monitor, SportsMOT, Agricultural deep-learning survey, Florence-2, scene graph survey, autonomous racing survey, remote sensing transformer survey, Visual-RFT, anomaly detection survey, GenEval.

Anchor-free boxes on a timeline

Temporal action localisation finds the start and end of an action in a video, so it is detection along time. AFSD is anchor-free: a 1D pyramid predicts the distance to the start and end of an action at each time step, and a boundary-pooling refinement takes the most salient boundary features of each proposal 2103.13137 §3. Author-reported, it reaches 52.0 average mAP on THUMOS14 with I3D features and 34.4 on ActivityNet 1.3 2103.13137 tab. 1. The table mixes backbones (I3D, TS, C3D, P3D), so compare rows within one backbone.

Detection and density for crowd counting

DecideNet counts people by blending two maps: a detector's head boxes turned into Gaussian bumps, and a five-layer regression map. A small attention network picks the blend pixel by pixel 1712.06679 §3.4 1712.06679 eq. 4. The authors report the lowest MAE of the compared methods on Mall, 1.52 against 1.82 for Exemplary Density [arxiv:1712.06679 tab. 1], and an average MAE of 9.23 on WorldExpo'10 1712.06679 tab. 3. On ShanghaiTech PartB it reports 21.53, while CP-CNN reports 20.1 1712.06679 tab. 2. In the authors' ablation, the regression map alone gives 3.37 MAE on Mall and the detection map alone 4.50 1712.06679 tab. 4. The Gaussian width of the detector map and the loss weight λ are not stated 1712.06679 §3.3 1712.06679 §4. Neither training nor inference cost is reported 1712.06679 §5.1.

Card: DecideNet.

Evaluation

The Metrics Reloaded consensus gives detection-specific rules for choosing a score 2206.01653 §ObD:

• Accuracy, specificity and AUROC are not valid for detection, because there is no meaningful count of true negatives. Use the three remaining entries of the confusion matrix 2206.01653 §ObD.
• With no continuous class scores, a counting metric such as F-beta is the default. With scores, the paper pairs a counting metric with AP or FROC 2206.01653 §ObD.
• The localisation criterion and the assignment strategy are chosen by decision guides in the supplement, which this page does not transcribe 2206.01653 §ObD [supplement: decision guides].

The paper reports no benchmark numbers. Its rules come from a multi-round expert consensus, with each component above 75% agreement 2206.01653 §Step 2.

For 3D detection on KITTI, the cards mix two AP conventions. Test-server results use 40 recall positions, and the validation tables use 11 2104.10330 §4.2 2104.10330 tab. 2. The 3D survey's case study also uses 11 positions on validation, while its headline test table uses 40 2106.10823 §4.2.1 2106.10823 tab. 6. Numbers from the two conventions should not be placed in one table.

The KITTI 11-point AP has a flaw that inflates scores at the low end. Precision at recall 0 is counted, so one correct detection gives 1/11, about 9.09% AP, per difficulty level 1905.12365 §6 1905.12365 tab. 5. The MonoDIS authors switch to 40 recall points, which drops recall 0 1905.12365 §6. Their test AP40 values were computed from published precision-recall values, not from the server 1905.12365 §7.4. Use AP40 only with other AP40 numbers.

For masks, Structure-measure scores a non-binary map as an average of two terms with equal weight, S = 0.5·S_o + 0.5·S_r 1708.00786 §4.3. The region term compares four blocks, split at the ground-truth centroid, with SSIM. The object term compares the mean and spread of the map inside and outside the object 1708.00786 §4.1 1708.00786 §4.2. In the authors' user study, 63.69% of trials preferred the top map by this measure over the one from weighted F-measure 1708.00786 §5.6. The official code weights the blocks by image area, not by ground-truth area as the paper describes, so a port has to state its choice S_region.m.

Open-world detection uses two error measures next to known-class mAP. Wilderness Impact (WI) and Absolute Open-Set Error (A-OSE) both count how often unknown objects are predicted as known classes, and lower is better 2103.02603 §5.1. WI cannot be measured on the last task, because no unknown boxes remain 2103.02603 §5.3.

When to use which

Implementation sketch

The shared core of a 2D detector, with the points where the variants differ. Same style as paper cards: Python-flavoured, lines of at most 80 characters.

# Common core of 2D object detection  [arxiv:1805.04953 §3] [arxiv:2211.12860 §3.1]
# Input:  image (H, W, 3); model with backbone, head and post-processing
# Output: list of (class, box, score)

def detect(image, model, score_threshold):
    # Shared convolutional features  [arxiv:1805.04953 §3]
    feats = model.backbone(image)
    # Candidates: anchors (SSD, YOLO), proposals (Faster R-CNN) or
    # learned queries (DETR-style)  [arxiv:1802.09567 §4.1] [arxiv:2211.12860 §3.1]
    candidates = model.make_candidates(feats)
    # Each candidate gets class scores and box offsets  [arxiv:1805.04953 §3.3]
    scores, boxes = model.head(feats, candidates)
    # Keep confident candidates  [arxiv:1802.09567 §4.1]
    keep = scores.max(axis=-1) > score_threshold
    scores, boxes = scores[keep], boxes[keep]
    # Dense and two-stage detectors drop duplicates with NMS, IoU 0.2 in RGB-N
    # [arxiv:1805.04953 §3.4]. Base DETR-style models skip this step
    # [arxiv:2211.12860 §1]
    if model.needs_nms:
        boxes, scores = nms(boxes, scores, iou_threshold=0.2)
    return boxes, scores

Contradictions

• No shared benchmark. The cards use different datasets and metrics: per-class recall, precision and an undefined "accuracy" on road damage 1801.09454 §5, plate read rate on frames 1802.09567 §5, pixel F1 on forgery benchmarks 1805.04953 §4.2, COCO AP for DETR-style models 2211.12860 §4.2 and 2011.10881 tab. 1, AP on AI-TOD 2208.08738 tab. 5, ScanNet and SUN RGB-D mAP 2104.00678 tab. 1, mAP on nuScenes 2205.13790 tab. 1, Wilderness Impact and A-OSE for open-world detection 2103.02603 §5.1, whole-scene mAP at IoU 0.25 on ARKitScenes 2111.08897 tab. 2, and AP on simulated LiDAR 2109.07644 §5.2, and for 3D, KITTI at 11 or 40 recall positions, nuScenes NDS and mAP, and Waymo APH 2104.10330 §4.2 2110.06922 tab. 1 2203.17270 tab. 3. Their numbers are not comparable across papers.
• Why DETR converges slowly. TSP attributes the slow training mainly to the cross-attention module 2011.10881 §1. H-DETR attributes it to few positive queries per object under one-to-one matching 2207.13080 §1. Both fixes help, but the cards do not test each cause on its own.
• Road-damage headline. The authors say SSD MobileNet is better overall and that the best class exceeds 0.75 recall and precision at 1.5 s on a phone 1801.09454 §5 1801.09454 §6. The table shows Inception V2 is better on several classes, and only D43 with Inception V2 clears 0.75 on both 1801.09454 tab. 2.
• Best polyp detector. The authors name YOLOv4 as the best detector 2011.07631 §VI. Their table gives RetinaNet with ResNet-101 the highest AP, 0.8745 against 0.8513 for YOLOv4 2011.07631 tab. III. The authors weigh speed as well as AP.
• ColonSegNet detection. The card reports AP 0.8000 at 180 FPS for ColonSegNet in detection, but the public repository implements only segmentation 2011.07631 tab. III README.md. The detection numbers cannot be reproduced from the release.
• NMS-free claim. The base Co-DETR method needs no NMS 2211.12860 §1. Its strongest Co-DINO results do use NMS 2211.12860 §4.3, so gains in that comparison mix the training scheme with post-processing.
• Which YOLO version is best. YOLO Evolution puts YOLO11 and YOLOv12 level on accuracy and YOLO11 first on balance. YOLO11 and YOLOv8 on orchard fruit finds YOLO11m-seg most accurate and YOLOv8n-seg fastest. The datasets, tasks and hardware differ, and each card reports one run per model 2411.00201 tab. 4 2410.19869 tab. 2 2410.19869 tab. 3.
• Text and figures in the YOLO benchmark. YOLO Evolution gives YOLOv5ul's traffic-sign mAP50 as 0.866 in the text and 0.886 in its figure [arxiv:2411.00201 fig. 12]. Its text says the tables use the test split, while the authors' code reports validation metrics, so the split behind the numbers is unclear 2411.00201 §3.1.
• Small-object gains. YOLO-Z's abstract gives up to 6.9% relative gain on small objects, while its Table 2 gives 5.46% to 7.73% 2112.11798 tab. 2. Only the table is used on this page.
• ReDet's abstract against its tables. The abstract gives gains of 1.2, 3.5 and 2.6 mAP over the previous best. Against the strongest other row in its tables the gains are 0.68, 1.90 and 2.62, and on HRSC2016 VOC2007 only 0.29 2103.07733 §1 2103.07733 tab. 6 2103.07733 tab. 8. The abstract's model sizes, 313 Mb and 121 Mb, appear in no table row 2103.07733 §1.
• Dynamic Head's cost. The abstract says the head adds no computational overhead. Table 2 shows +21.50 GFLOPs at six blocks, and the text calls this negligible 2106.08322 tab. 2 2106.08322 §4.2. Its 54.0 AP result uses test-time augmentation and a DCN backbone, so the fairer comparison with HTC++ is its 58.7 AP without extra data [arxiv:2106.08322 §4.3] 2106.08322 tab. 7.
• Summary table against the text, YOLO review. The review's summary table and its prose give different AP for several versions, and some table entries come from other settings. The table gives 33.0 for YOLOv3, where the text gives 36.2 for YOLOv3-spp 2304.00501 tab. 4 2304.00501 §6.3. The table's 55.8 for YOLOv5 is the 1536-pixel figure with test-time augmentation, while the 640-pixel figure in the text is 50.7 2304.00501 §9. Use the original papers for exact values.
• Schedules in the backbone comparison. The PVT v2 text says its comparisons use 1x or 3x schedules, but its Table 4 does not say which one was used for ATSS, GFL or Cascade Mask R-CNN 2106.13797 §4.2 [arxiv:2106.13797 tab. 4]. Compare those rows only with that caveat.
• PVT v2 learning rate. The text gives 1e-4 for Mask R-CNN at 1x, while the released 1x config sets 2e-4 2106.13797 §4.2 [code: detection/configs/mask_rcnn_pvt_v2_b2_fpn_1x_coco.py#L20]. Pick one and state which.
• RFAConv's replacement scope. The paper says every 3×3 convolution in the YOLOv5 and YOLOv8 yaml files is replaced 2304.03198 §4.3. The shipped yaml files replace only some of them [code: Detection/YOLOv5/models/yolov5n.yaml (NovelConv)#L16-L22] YOLOv8n.yaml, and the NovelConv class they use is not in the repository. The detection numbers therefore cannot be reproduced from that code 2304.03198 §4.3.
• DETR3D's depth claim. The authors say DETR3D's mATE is worse than FCOS3D's, and attribute the gap to FCOS3D's depth supervision 2110.06922 §4.2. The card records this as the authors' suspicion, not a measured cause.
• Few-shot headline. The text says FSCE gains up to 8.8 nAP50 on split 3 at 10 shots. That gain is measured against MPSR, at 49.7. TFA w/ fc is 50.2 in the same column, so the margin to the best unstarred row is 8.3 [arxiv:2103.05950 §4.2] 2103.05950 tab. 2.
• COD10K test size. The COD-Rank card gives COD10K 2,026 test images 2103.04011 §1 2103.04011 §2. SINet-V2 gives 4,000 test images in its split 2102.10274 §3.3 2102.10274 tab. 1, and 2,026 in its results 2102.10274 §5.2. Compare COD10K numbers only with the split stated.
• MARTENS failure counts. The results tables give 50% (use case 1) and 26% (use case 2) more failures than random search 2410.19277 tab. 2 2410.19277 tab. 3. The paper's abstract and conclusion state the gain as a range (25% to 50%, or 26% to 50%), and the printed row sums do not match the printed totals 2410.19277 tab. 2. Both figures are shown; they are not reconciled here.
• Edge YOLO's model size. The abstract gives 25.67 MB, while Table 5 gives 25.27 MB 2205.14942 tab. 5. The table value is used on this page.
• Edge YOLO's KITTI mAP. The text gives KITTI mAP values of 0.7904 and 0.8473 2205.14942 §IV-C, while Table 6 lists YOLOv3-Tiny at 0.7804 2205.14942 tab. 6. The table rows are used on this page.
• Survey text against its own table. The 3D survey gives PV-RCNN++ an easy AP3D of 90.90 in its text, while its table gives PV-RCNN++ 90.14 and gives 90.90 to Voxel R-CNN 2206.09474 §10.1.1 2206.09474 tab. 16.

Papers

• Road Damage Detection Using Deep Neural Networks with Images Captured Through a Smartphone: SSD baselines on a public road-damage dataset, with phone and GPU timings.
• FSSD: Feature Fusion Single Shot Multibox Detector: fuses three SSD source layers into one map before the detection pyramid.
• xView: Objects in Context in Overhead Imagery: an overhead detection dataset with an SSD baseline and three training-set variants.
• Real-time ALPR with YOLO (Laroca et al.): a cascade of YOLO-style detectors for plates, with per-stage timing.
• Real-Time Polyp Detection, Localization and Segmentation in Colonoscopy Using Deep Learning: benchmark of detectors and segmenters on a colonoscopy dataset, with a fast proposed network.
• Rethinking Transformer-based Set Prediction for Object Detection: a transformer encoder over selected features, trained with a restricted matching loss.
• End-to-end Interpretable Neural Motion Planner: one network for LiDAR-based detection, motion and a planning cost map.
• Unbiased Teacher for Semi-Supervised Object Detection: teacher-student pseudo-labelling for Faster R-CNN with few labels.
• ICDAR2019 Competition on Scanned Receipt OCR and Information Extraction: text localisation, OCR and field extraction on receipts, with the top entries.
• Group-Free 3D Object Detection via Transformers: transformer decoder over all points for 3D indoor detection, with no point grouping.
• A Comprehensive Survey of Scene Graphs: Generation and Application: surveys relation detection on top of object detection, with reported benchmark numbers.
• SeaDronesSee: A Maritime Benchmark for Detecting Humans in Open Water: drone dataset for detecting and tracking people and boats in open water.
• OPV2V: a simulated benchmark and intermediate-fusion pipeline for 3D detection between cars.
• Autonomous Vehicles on the Edge: A Survey on Autonomous Vehicle Racing: survey of perception, planning and control for racing cars, with open problems.
• Knowledge Distillation from A Stronger Teacher: Pearson-correlation distillation, with COCO detection results.
• BEVFusion: A Simple and Robust LiDAR-Camera Fusion Framework: camera and LiDAR streams in a shared BEV grid, robust to LiDAR loss.
• DETRs with Hybrid Matching: a training-only one-to-many group added to DETR-style detectors.
• RFLA: Gaussian Receptive Field based Label Assignment for Tiny Object Detection: label assignment by receptive-field distance, for tiny objects.
• Transformers in Remote Sensing: A Survey: survey of transformer methods for remote sensing, with reported detection numbers.
• Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks: one model for captioning, detection and grounding, written as text.
• DETRs with Collaborative Hybrid Assignments Training: auxiliary one-to-many heads that speed up DETR-style training, with the inference path unchanged.
• YOLO-LITE: A Real-Time Object Detection Algorithm Optimized for Non-GPU Computers: a small detector found from Tiny-YOLOv2, for laptops and phones without a GPU.
• microYOLO: Towards Single-Shot Object Detection on Microcontrollers: YOLO-style detector for a Cortex-M7 board, with pruning and 8-bit quantization.
• ARKitScenes: A Diverse Real-World Dataset For 3D Indoor Scene Understanding Using Mobile RGB-D Data: indoor RGB-D captures with 3D furniture boxes and three detector baselines.
• Anomaly Detection in Autonomous Driving: A Survey: survey of corner-case detection across camera, lidar, radar and object-level data.
• Metrics reloaded: Recommendations for image analysis validation: expert-consensus decision guide for choosing validation metrics, including detection.
• RDD2022: A multi-national image dataset for automatic Road Damage Detection: six-country road-damage dataset with four classes and no detector baseline.
• OrientedFormer: DETR-style detector for rotated boxes in aerial images.
• Visual-RFT: Visual Reinforcement Fine-Tuning: reinforcement fine-tuning of a vision-language model with a box reward.
• Synscapes: A Photorealistic Synthetic Dataset for Street Scene Parsing: synthetic street scenes with instance, depth and box labels.
• Monitoring COVID-19 social distancing with person detection and tracking via fine-tuned YOLO v3 and Deepsort techniques: YOLO v3 with DeepSORT for a distance monitor, with unvalidated grouping.
• Equalization Loss v2: A New Gradient Balance Approach for Long-tailed Object Detection: per-class gradient reweighting for sigmoid detectors, in one stage.
• YOLO5Face: Why Reinventing a Face Detector: YOLOv5 adapted to faces, with a landmark head and small mobile models.
• Disentangle Your Dense Object Detector: separate assignment, deformable head layers and level weights for one-stage detectors.
• PubTables-1M: Towards comprehensive table extraction from unstructured documents: table dataset with DETR baselines for table structure.
• YOLO-Z: Improving small object detection in YOLOv5 for autonomous vehicles: YOLOv5 changes for small traffic cones, with a higher-resolution map.
• PolarFormer: Multi-camera 3D Object Detection with Polar Transformer: camera-only 3D detection on a polar bird's-eye-view grid.
• DeepInteraction: 3D Object Detection via Modality Interaction: LiDAR and camera features kept apart and exchanged during detection.
• SportsMOT: A Large Multi-Object Tracking Dataset in Multiple Sports Scenes: tracking dataset for players in three sports, with a tracker that adds appearance.
• What is YOLOv8: An In-Depth Exploration of the Internal Features of the Next-Generation Object Detector: descriptive review of YOLOv8, with no experiments of its own.
• Comparing YOLOv11 and YOLOv8 for instance segmentation of occluded and non-occluded immature green fruits in complex orchard environment: YOLO11-seg and YOLOv8-seg benchmark on orchard apples.
• YOLO Evolution: A Comprehensive Benchmark and Architectural Review of YOLOv12, YOLO11, and Their Previous Versions: 33 YOLO checkpoints on three datasets, one training setting.
• Joint 3D Proposal Generation and Object Detection from View Aggregation: LiDAR and RGB fused into 3D car boxes through a proposal network.
• Deep learning in agriculture: A survey: survey of 40 agricultural deep-learning studies, with tabulated settings and results.
• ReDet: A Rotation-equivariant Detector for Aerial Object Detection: rotation-equivariant backbone and RiRoI Align for oriented aerial boxes.
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows: windowed transformer backbone, used with detectors on COCO.
• BADet: Boundary-Aware 3D Object Detection from Point Clouds: graph refinement of LiDAR proposals for 3D detection.
• Dynamic Head: Unifying Object Detection Heads with Attentions: stacked scale, spatial and task attention blocks as a detector head.
• 3D Object Detection for Autonomous Driving: A Survey: survey of 3D detection by input type, with a case study of fifteen detectors.
• DETR3D: 3D Object Detection from Multi-view Images via 3D-to-2D Queries: camera-only 3D detection from learned object queries, with no NMS.
• BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers: BEV grid from surround cameras, with spatial and temporal attention.
• A Comprehensive Review of YOLO Architectures in Computer Vision: From YOLOv1 to YOLOv8 and YOLO-NAS: review of YOLO versions with a summary table of reported AP.
• YOLOv10: Real-Time End-to-End Object Detection: YOLO without NMS, with one-to-many and one-to-one heads.
• YOLOv11: An Overview of the Key Architectural Enhancements: architectural overview of YOLOv11, with no experiments of its own.
• A Survey of Modern Deep Learning based Object Detection Models: survey of detector families, datasets and metrics, with compiled author-reported numbers.
• Conformer: Local Features Coupling Global Representations for Visual Recognition: hybrid CNN and transformer backbone, with COCO detection results.
• GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment: text-to-image benchmark whose checks rely on a COCO-trained detector.
• Categorical Depth Distribution Network for Monocular 3D Object Detection: camera-only car detector that lifts per-pixel depth distributions into bird's-eye view.
• Voxel Transformer for 3D Object Detection: Transformer backbone over non-empty voxels for LiDAR detection, with a hash-table neighbour search.
• TopoLogic: An Interpretable Pipeline for Lane Topology Reasoning on Driving Scenes: lane-to-lane connection scores from endpoint gaps and query similarity, fused and passed to a graph layer.
• Learning Rich Features for Image Manipulation Detection: Faster R-CNN with RGB and noise streams for forgery localisation.
• PVT v2: Improved Baselines with Pyramid Vision Transformer: pyramid transformer backbone with pooled-key attention, overlapping patches and a depthwise convolution in the feed-forward block.
• RFAConv: Receptive-Field Attention Convolution for Improving Convolutional Neural Networks: convolution with per-window attention over each receptive field, as a drop-in replacement.
• Towards Open World Object Detection: Faster R-CNN variant that flags unknown objects and learns them incrementally.
• RTMDet: An Empirical Study of Designing Real-Time Object Detectors: one-stage detector family from tiny to extra-large, with a 5 by 5 depthwise block and soft label assignment.
• VisionLLM: Large Language Model is also an Open-Ended Decoder for Vision-Centric Tasks: detection and other vision tasks written as text and decoded by an LLM.
• POI: Multiple Object Tracking with High Performance Detection and Appearance Feature: Faster R-CNN detections with a learned appearance feature for tracking.
• Video Salient Object Detection via Fully Convolutional Networks: static and dynamic FCNs on frame pairs, trained on warped still images.
• Deep Visual Attention Prediction: eye-fixation model with three deconvolution side outputs and a learned fusion.
• Structure-measure: A New Way to Evaluate Foreground Maps: region and object structural similarity score for non-binary maps.
• Disentangling Monocular 3D Object Detection: monocular car detector with a split 3D loss and a 40-point AP variant.
• Rethinking RGB-D Salient Object Detection: Models, Data Sets, and Large-Scale Benchmarks: three-stream RGB-D model with a test-time depth gate and a phone dataset.
• EGNet: Edge Guidance Network for Salient Object Detection: salient-edge path that guides each object scale of a salient-object network.
• Concealed Object Detection: SINet-V2 for camouflaged objects, with the COD10K dataset.
• Simultaneously Localize, Segment and Rank the Camouflaged Objects: fixation, camouflage mask and difficulty rank in one network; ranking code absent.
• FSCE: Few-Shot Object Detection via Contrastive Proposal Encoding: contrastive head on RoI proposals for few-shot fine-tuning.
• Anabranch Network for Camouflaged Object Segmentation: classifier stream that scales the map of a saliency segmenter; CAMO dataset.
• YOLOP: You Only Look Once for Panoptic Driving Perception: one encoder with heads for vehicles, drivable area and lane lines.
• YOLO-Pose: Enhancing YOLO for Multi Person Pose Estimation Using Object Keypoint Similarity Loss: YOLO head with keypoints per anchor and an OKS-based keypoint loss.
• MARTENS: genetic search for failing arm scenes in simulation, then detector repair and real-image fine-tuning.
• Moving Forward: survey and benchmark of autonomous-driving software and hardware, with a detection results table.
• Pseudo-LiDAR++: Accurate Depth for 3D Object Detection in Autonomous Driving: stereo depth trained on depth, corrected with sparse LiDAR points by graph propagation.
• CenterFusion: Center-based Radar and Camera Fusion for 3D Object Detection: radar points matched to camera detections and painted as depth and velocity maps.
• Inner-IoU: box regression loss computed on scaled auxiliary boxes, with no added term.
• DecideNet (crowd counting): detector boxes as Gaussian bumps, blended with a regression map by a per-pixel attention net.
• LaserNet: An Efficient Probabilistic 3D Object Detector for Autonomous Driving: range-image LiDAR detector with a mixture of box distributions and variance-aware NMS.
• A Review on Deep Learning in UAV Remote Sensing: literature review of deep learning on drone imagery, tallied by network, task and sensor.
• A Survey of Deep Learning Techniques for Weed Detection from Images: survey of 70 deep-learning weed-detection papers and 19 public datasets.
• Edge YOLO: Real-Time Intelligent Object Detection System Based on Edge-Cloud Cooperation in Autonomous Vehicles: slimmed YOLOv4 for an in-vehicle Jetson board, retrained in the cloud.
• 3D Object Detection for Autonomous Driving: A Comprehensive Survey: survey of 3D detection by sensor, with compiled author-reported scores.
• YOLO advances to its genesis: a decadal and comprehensive review of the You Only Look Once (YOLO) series: review of YOLOv1 to YOLOv12 and variant detectors, with no experiments
• Learning Salient Boundary Feature for Anchor-free Temporal Action Localization: anchor-free temporal action localiser with boundary pooling.
• Image Data Augmentation for Deep Learning: A Survey: augmentation survey with its own detection, segmentation and classification runs.
• V2V4Real: A Real-world Large-scale Dataset for Vehicle-to-Vehicle Cooperative Perception: two-car LiDAR dataset with cooperative detection, tracking and Sim2Real benchmarks. of its own.