Metric: APS (higher is better)
| # | Model↕ | APS▼ | Augmentations | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | Focal-Stable-DINO (Focal-Huge, no TTA) | 48.6 | No | A Strong and Reproducible Object Detector with O... | 2023-04-25 | Code |
| 2 | EVA | 48.5 | No | EVA: Exploring the Limits of Masked Visual Repre... | 2022-11-14 | Code |
| 3 | Group DETR v2 | 48.4 | No | Group DETR v2: Strong Object Detector with Encod... | 2022-11-07 | - |
| 4 | Plain-DETR (Swin-L) | 48.2 | No | - | - | Code |
| 5 | Relation-DETR (Focal-L) | 47.2 | No | Relation DETR: Exploring Explicit Position Relat... | 2024-07-16 | Code |
| 6 | DETA (Swin-L) | 46.1 | No | NMS Strikes Back | 2022-12-12 | Code |
| 7 | GLIP (Swin-L, multi-scale) | 45.3 | No | Grounded Language-Image Pre-training | 2021-12-07 | Code |
| 8 | PyCenterNet (Swin-L, multi-scale) | 38.7 | No | CenterNet++ for Object Detection | 2022-04-18 | Code |
| 9 | CenterNet2 (Res2Net-101-DCN-BiFPN, self-training, 1560 single-scale) | 38.7 | No | Probabilistic two-stage detection | 2021-03-12 | Code |
| 10 | DetectoRS (ResNeXt-101-64x4d, multi-scale) | 37.7 | No | DetectoRS: Detecting Objects with Recursive Feat... | 2020-06-03 | Code |
| 11 | SOLQ (Swin-L, single scale) | 37.6 | No | SOLQ: Segmenting Objects by Learning Queries | 2021-06-04 | Code |
| 12 | QueryInst (single-scale) | 37.4 | No | Instances as Queries | 2021-05-05 | Code |
| 13 | DetectoRS (ResNeXt-101-32x4d, multi-scale) | 37.4 | No | DetectoRS: Detecting Objects with Recursive Feat... | 2020-06-03 | Code |
| 14 | YOLOv4-P6 CSP-P6 (single-scale, 32 fps) | 36.6 | No | Scaled-YOLOv4: Scaling Cross Stage Partial Network | 2020-11-16 | Code |
| 15 | PAA (ResNext-152-32x8d + DCN, multi-scale) | 36 | No | Probabilistic Anchor Assignment with IoU Predict... | 2020-07-16 | Code |
| 16 | UniverseNet-20.08d (Res2Net-101, DCN, multi-scale) | 35.8 | No | USB: Universal-Scale Object Detection Benchmark | 2021-03-25 | Code |
| 17 | Mask R-CNN (ResNet-101-FPN, CBN) | 35.8 | No | Cross-Iteration Batch Normalization | 2020-02-13 | Code |
| 18 | GFLV2 (Res2Net-101, DCN, multiscale) | 35.7 | No | Generalized Focal Loss V2: Learning Reliable Loc... | 2020-11-25 | Code |
| 19 | Cascade Mask R-CNN (Triple-ResNeXt152, multi-scale) | 35.5 | No | CBNet: A Novel Composite Backbone Network Archit... | 2019-09-09 | Code |
| 20 | RetinaNet (SpineNet-190, 1280x1280) | 35.4 | No | SpineNet: Learning Scale-Permuted Backbone for R... | 2019-12-10 | Code |
| 21 | LSNet (Res2Net-101+ DCN, multi-scale) | 35.2 | No | Location-Sensitive Visual Recognition with Cross... | 2021-04-11 | Code |
| 22 | ResNeSt-200 (multi-scale) | 35.1 | No | ResNeSt: Split-Attention Networks | 2020-04-19 | Code |
| 23 | RepPoints v2 (ResNeXt-101, DCN, multi-scale) | 34.5 | No | RepPoints V2: Verification Meets Regression for ... | 2020-07-16 | Code |
| 24 | Deformable DETR (ResNeXt-101+DCN) | 34.4 | No | Deformable DETR: Deformable Transformers for End... | 2020-10-08 | Code |
| 25 | AC-FPN Cascade R-CNN (X-152-32x8d-FPN-IN5k, multi scale, only CEM) | 34.2 | No | Attention-guided Context Feature Pyramid Network... | 2020-05-23 | Code |
| 26 | NAS-FPN (AmoebaNet-D, learned aug) | 34.2 | No | Learning Data Augmentation Strategies for Object... | 2019-06-26 | Code |
| 27 | OTA (ResNeXt-101+DCN, multiscale) | 34.1 | No | OTA: Optimal Transport Assignment for Object Det... | 2021-03-26 | Code |
| 28 | DetectoRS (ResNeXt-101-32x4d, single-scale) | 33.9 | No | DetectoRS: Detecting Objects with Recursive Feat... | 2020-06-03 | Code |
| 29 | TSD(SENet154-DCN,multi-scale) | 33.8 | No | Revisiting the Sibling Head in Object Detector | 2020-03-17 | Code |
| 30 | RetinaNet (SpineNet-143, 1280x1280) | 33.6 | No | SpineNet: Learning Scale-Permuted Backbone for R... | 2019-12-10 | Code |
| 31 | PP-YOLOE-x(CSPRepResNet-x, 640x640, single-scale ) | 33.3 | No | PP-YOLOE: An evolved version of YOLO | 2022-03-30 | Code |
| 32 | ATSS (ResNetXt-64x4d-101+DCN,multi-scale) | 33.2 | No | Bridging the Gap Between Anchor-based and Anchor... | 2019-12-05 | Code |
| 33 | Dynamic R-CNN (ResNet-101-DCN, multi-scale) | 32.8 | No | Dynamic R-CNN: Towards High Quality Object Detec... | 2020-04-13 | Code |
| 34 | D2Det (ResNet-101-DCN, multi-scale test) | 32.7 | No | - | - | Code |
| 35 | TSD(ResNet-101-Deformable, Image Pyramid) | 32.7 | No | Revisiting the Sibling Head in Object Detector | 2020-03-17 | Code |
| 36 | DAT-S (RetinaNet) | 32.3 | No | Vision Transformer with Deformable Attention | 2022-01-03 | Code |
| 37 | aLRP Loss (ResNext-101-64x4d, DCN, multiscale test) | 32 | No | A Ranking-based, Balanced Loss Function Unifying... | 2020-09-28 | Code |
| 38 | RetinaNet (SpineNet-96, 1024x1024) | 32 | No | SpineNet: Learning Scale-Permuted Backbone for R... | 2019-12-10 | Code |
| 39 | TridentNet (ResNet-101-Deformable, Image Pyramid) | 31.8 | No | Scale-Aware Trident Networks for Object Detection | 2019-01-07 | Code |
| 40 | UniverseNet-20.08d (Res2Net-101, DCN, single-scale) | 31.7 | No | USB: Universal-Scale Object Detection Benchmark | 2021-03-25 | Code |
| 41 | PP-YOLOE-l(CSPRepResNet-l, 640x640, single-scale ) | 31.4 | No | PP-YOLOE: An evolved version of YOLO | 2022-03-30 | Code |
| 42 | PPDet (ResNeXt-101-FPN, multiscale) | 31.4 | No | Reducing Label Noise in Anchor-Free Object Detec... | 2020-08-03 | Code |
| 43 | GFLV2 (Res2Net-101, DCN) | 31.3 | No | Generalized Focal Loss V2: Learning Reliable Loc... | 2020-11-25 | Code |
| 44 | FreeAnchor + SEPC (DCN, ResNext-101-64x4d) | 31.3 | No | Scale-Equalizing Pyramid Convolution for Object ... | 2020-05-06 | Code |
| 45 | YOLOX-X (Modified CSP v5) | 31.2 | No | YOLOX: Exceeding YOLO Series in 2021 | 2021-07-18 | Code |
| 46 | CPNDet (Hourglass-104, multi-scale) | 31 | No | Corner Proposal Network for Anchor-free, Two-sta... | 2020-07-27 | Code |
| 47 | aLRP Loss (ResNext-101-64x4d, DCN, single scale) | 30.8 | No | A Ranking-based, Balanced Loss Function Unifying... | 2020-09-28 | Code |
| 48 | RepPoints v2 (ResNeXt-101, DCN) | 30.3 | No | RepPoints V2: Verification Meets Regression for ... | 2020-07-16 | Code |
| 49 | RPDet (ResNet-101-DCN, multi-scale) | 30.3 | No | RepPoints: Point Set Representation for Object D... | 2019-04-25 | Code |
| 50 | aLRP Loss (ResNext-101-64x4d, single scale) | 30.2 | No | A Ranking-based, Balanced Loss Function Unifying... | 2020-09-28 | Code |
| 51 | UniverseNet-20.08 (Res2Net-50, DCN, single-scale) | 30.1 | No | USB: Universal-Scale Object Detection Benchmark | 2021-03-25 | Code |
| 52 | PANet (ResNeXt-101, multi-scale) | 30.1 | No | Path Aggregation Network for Instance Segmentation | 2018-03-05 | Code |
| 53 | GFLV2 (ResNeXt-101, 32x4d, DCN) | 29.7 | No | Generalized Focal Loss V2: Learning Reliable Loc... | 2020-11-25 | Code |
| 54 | MatrixNet Corners (ResNet-152, multi-scale) | 29.7 | No | Matrix Nets: A New Deep Architecture for Object ... | 2019-08-13 | Code |
| 55 | FSAF (ResNeXt-101, multi-scale) | 29.7 | No | Feature Selective Anchor-Free Module for Single-... | 2019-03-02 | Code |
| 56 | SNIPER (ResNet-101) | 29.6 | No | SNIPER: Efficient Multi-Scale Training | 2018-05-23 | Code |
| 57 | M2Det (ResNet-101, multi-scale) | 29.6 | No | M2Det: A Single-Shot Object Detector based on Mu... | 2018-11-12 | Code |
| 58 | D-RFCN + SNIP (DPN-98 with flip, multi-scale) | 29.3 | No | An Analysis of Scale Invariance in Object Detect... | 2017-11-22 | - |
| 59 | GFL (X-101-32x4d-DCN, single-scale) | 29.2 | No | Generalized Focal Loss: Learning Qualified and D... | 2020-06-08 | Code |
| 60 | M2Det (VGG-16, multi-scale) | 29.2 | No | M2Det: A Single-Shot Object Detector based on Mu... | 2018-11-12 | Code |
| 61 | RetinaNet (SpineNet-49, 896x896) | 29.1 | No | SpineNet: Learning Scale-Permuted Backbone for R... | 2019-12-10 | Code |
| 62 | HoughNet (MS) | 29.1 | No | HoughNet: Integrating near and long-range eviden... | 2020-07-05 | Code |
| 63 | CenterNet511 (Hourglass-104, multi-scale) | 28.9 | No | CenterNet: Keypoint Triplets for Object Detection | 2019-04-17 | Code |
| 64 | GFLV2 (ResNet-101-DCN) | 28.8 | No | Generalized Focal Loss V2: Learning Reliable Loc... | 2020-11-25 | Code |
| 65 | ISTR (ResNet101-FPN-3x, single-scale) | 28.7 | No | ISTR: End-to-End Instance Segmentation with Tran... | 2021-05-03 | Code |
| 66 | PP-YOLOE-m(CSPRepResNet-m, 640x640, single-scale ) | 28.6 | No | PP-YOLOE: An evolved version of YOLO | 2022-03-30 | Code |
| 67 | SAPD (ResNeXt-101, single-scale) | 28.1 | No | Soft Anchor-Point Object Detection | 2019-11-27 | Code |
| 68 | HTC (HRNetV2p-W48) | 28 | No | Deep High-Resolution Representation Learning for... | 2019-08-20 | Code |
| 69 | ISTR (ResNet50-FPN-3x, single-scale) | 27.8 | No | ISTR: End-to-End Instance Segmentation with Tran... | 2021-05-03 | Code |
| 70 | GFLV2 (ResNet-101) | 27.8 | No | Generalized Focal Loss V2: Learning Reliable Loc... | 2020-11-25 | Code |
| 71 | DCNv2 (ResNet-101, multi-scale) | 27.8 | No | Deformable ConvNets v2: More Deformable, Better ... | 2018-11-27 | Code |
| 72 | CenterMask+VoVNetV2-99 (single-scale) | 27.8 | No | CenterMask : Real-Time Anchor-Free Instance Segm... | 2019-11-15 | Code |
| 73 | FCOS (ResNeXt-64x4d-101-FPN 4 + improvements) | 27.6 | No | FCOS: Fully Convolutional One-Stage Object Detec... | 2019-04-02 | Code |
| 74 | InterNet (ResNet-101-FPN, multi-scale) | 27.2 | No | Feature Intertwiner for Object Detection | 2019-03-28 | Code |
| 75 | D-RFCN + SNIP (ResNet-101, multi-scale) | 27.2 | No | An Analysis of Scale Invariance in Object Detect... | 2017-11-22 | - |
| 76 | Mask R-CNN (HRNetV2p-W48 + cascade) | 27.1 | No | Deep High-Resolution Representation Learning for... | 2019-08-20 | Code |
| 77 | CenterMask+VoVNet2-57 (single-scale) | 27.1 | No | CenterMask : Real-Time Anchor-Free Instance Segm... | 2019-11-15 | Code |
| 78 | YOLOv4 (CD53) | 27 | Yes | Scaled-YOLOv4: Scaling Cross Stage Partial Network | 2020-11-16 | Code |
| 79 | FreeAnchor (ResNeXt-101) | 27 | No | FreeAnchor: Learning to Match Anchors for Visual... | 2019-09-05 | Code |
| 80 | YOLOv3 @800 + ASFF* (Darknet-53) | 27 | Yes | Learning Spatial Fusion for Single-Shot Object D... | 2019-11-21 | Code |
| 81 | AC-FPN Cascade R-CNN(ResNet-101, single scale) | 26.9 | No | Attention-guided Context Feature Pyramid Network... | 2020-05-23 | Code |
| 82 | GFLV2 (ResNet-50) | 26.8 | No | Generalized Focal Loss V2: Learning Reliable Loc... | 2020-11-25 | Code |
| 83 | FoveaBox (ResNeXt-101) | 26.8 | No | FoveaBox: Beyond Anchor-based Object Detector | 2019-04-08 | Code |
| 84 | YOLOv4-608 | 26.7 | Yes | YOLOv4: Optimal Speed and Accuracy of Object Det... | 2020-04-23 | Code |
| 85 | FCOS (ResNeXt-101-64x4d-FPN) | 26.5 | No | FCOS: Fully Convolutional One-Stage Object Detec... | 2019-04-02 | Code |
| 86 | Cascade R-CNN-FPN (ResNet-101, map-guided) | 26.3 | No | InstaBoost: Boosting Instance Segmentation via P... | 2019-08-21 | Code |
| 87 | SNIPER (ResNet-50) | 26.1 | No | SNIPER: Efficient Multi-Scale Training | 2018-05-23 | Code |
| 88 | FCOS (ResNeXt-32x8d-101-FPN) | 26 | No | FCOS: Fully Convolutional One-Stage Object Detec... | 2019-04-02 | Code |
| 89 | RetinaNet (SpineNet-49, 640x640) | 25.9 | No | SpineNet: Learning Scale-Permuted Backbone for R... | 2019-12-10 | Code |
| 90 | RefineDet512+ (ResNet-101) | 25.6 | No | Single-Shot Refinement Neural Network for Object... | 2017-11-18 | Code |
| 91 | Faster R-CNN (LIP-ResNet-101-MD w FPN) | 25.4 | No | LIP: Local Importance-based Pooling | 2019-08-12 | Code |
| 92 | FCOS (HRNet-W32-5l) | 25.4 | No | FCOS: Fully Convolutional One-Stage Object Detec... | 2019-04-02 | Code |
| 93 | Libra R-CNN (ResNeXt-101-FPN) | 25.3 | No | Libra R-CNN: Towards Balanced Learning for Objec... | 2019-04-04 | Code |
| 94 | Grid R-CNN (ResNeXt-101-FPN) | 25.1 | No | Grid R-CNN | 2018-11-29 | Code |
| 95 | RPDet (ResNet-101-DCN) | 24.9 | No | RepPoints: Point Set Representation for Object D... | 2019-04-25 | Code |
| 96 | Faster R-CNN (HRNetV2p-W48) | 24.9 | No | Deep High-Resolution Representation Learning for... | 2019-08-20 | Code |
| 97 | RetinaMask (ResNeXt-101-FPN-GN) | 24.8 | No | RetinaMask: Learning to predict masks improves s... | 2019-01-10 | Code |
| 98 | aLRP Loss (ResNext-101, DCN, 500 scale) | 24.6 | No | A Ranking-based, Balanced Loss Function Unifying... | 2020-09-28 | Code |
| 99 | ResNet-50-DW-DPN (Deformable Kernels) | 24.6 | No | Deformable Kernels: Adapting Effective Receptive... | 2019-10-07 | Code |
| 100 | CornerNet-Saccade (Hourglass-104, multi-scale) | 24.4 | No | CornerNet-Lite: Efficient Keypoint Based Object ... | 2019-04-18 | Code |
| 101 | ExtremeNet (Hourglass-104, multi-scale) | 24.1 | No | Bottom-up Object Detection by Grouping Extreme a... | 2019-01-23 | Code |
| 102 | RetinaNet (ResNeXt-101-FPN) | 24.1 | No | Focal Loss for Dense Object Detection | 2017-08-07 | Code |
| 103 | YOLOF-DC5 | 24 | No | You Only Look One-level Feature | 2021-03-17 | Code |
| 104 | FSAF (ResNet-101, single-scale) | 24 | No | Feature Selective Anchor-Free Module for Single-... | 2019-03-02 | Code |
| 105 | TridentNet (ResNet-101) | 23.9 | No | Scale-Aware Trident Networks for Object Detection | 2019-01-07 | Code |
| 106 | SpineNet-49 (640, RetinaNet, single-scale) | 23.7 | No | SpineNet: Learning Scale-Permuted Backbone for R... | 2019-12-10 | Code |
| 107 | Cascade R-CNN (ResNet-101-FPN+, cascade) | 23.7 | No | Cascade R-CNN: Delving into High Quality Object ... | 2017-12-03 | Code |
| 108 | Cascade R-CNN | 23.7 | No | Cascade R-CNN: High Quality Object Detection and... | 2019-06-24 | Code |
| 109 | RPDet (ResNet-101) | 23.6 | No | RepPoints: Point Set Representation for Object D... | 2019-04-25 | Code |
| 110 | FCOS (HRNetV2p-W48) | 23.4 | Yes | Deep High-Resolution Representation Learning for... | 2019-08-20 | Code |
| 111 | RetinaNet (SpineNet-49S, 640x640) | 23.3 | No | SpineNet: Learning Scale-Permuted Backbone for R... | 2019-12-10 | Code |
| 112 | PP-YOLOE-s(CSPRepResNet-s, 640x640, single-scale ) | 23.2 | No | PP-YOLOE: An evolved version of YOLO | 2022-03-30 | Code |
| 113 | HTC (ResNeXt-101-FPN) | 22.8 | No | Hybrid Task Cascade for Instance Segmentation | 2019-01-22 | Code |
| 114 | HSD (Rest101, 768x768, single-scale test) | 22.8 | No | - | - | Code |
| 115 | RefineDet512+ (VGG-16) | 22.7 | No | Single-Shot Refinement Neural Network for Object... | 2017-11-18 | Code |
| 116 | Cascade R-CNN (ResNet-50-FPN+, cascade) | 22.6 | No | Cascade R-CNN: Delving into High Quality Object ... | 2017-12-03 | Code |
| 117 | GHM-C + GHM-R (RetinaNet-FPN-ResNeXt-101) | 22.3 | No | Gradient Harmonized Single-stage Detector | 2018-11-13 | Code |
| 118 | CenterNet (HRNetV2-W48) | 22.2 | No | Deep High-Resolution Representation Learning for... | 2019-08-20 | Code |
| 119 | M2Det (VGG-16, single-scale) | 22.1 | No | M2Det: A Single-Shot Object Detector based on Mu... | 2018-11-12 | Code |
| 120 | RDSNet (ResNet-101, RetinaNet, mask, MBRM) | 22.1 | No | RDSNet: A New Deep Architecture for Reciprocal O... | 2019-12-11 | Code |
| 121 | Fast R-CNN (Cascade RPN) | 22.1 | Yes | Cascade RPN: Delving into High-Quality Region Pr... | 2019-09-15 | Code |
| 122 | Mask R-CNN (ResNeXt-101-FPN) | 22.1 | No | Mask R-CNN | 2017-03-20 | Code |
| 123 | Faster R-CNN (Cascade RPN) | 22 | Yes | Cascade RPN: Delving into High-Quality Region Pr... | 2019-09-15 | Code |
| 124 | RetinaMask (ResNet-50-FPN) | 21.9 | No | RetinaMask: Learning to predict masks improves s... | 2019-01-10 | Code |
| 125 | GA-Faster-RCNN | 21.8 | No | Region Proposal by Guided Anchoring | 2019-01-10 | Code |
| 126 | RetinaNet (ResNet-101-FPN) | 21.8 | No | Focal Loss for Dense Object Detection | 2017-08-07 | Code |
| 127 | CenterNet-DLA (DLA-34, multi-scale) | 21.5 | No | Objects as Points | 2019-04-16 | Code |
| 128 | Cascade R-CNN (ResNet-101-FPN+) | 21.3 | No | Cascade R-CNN: Delving into High Quality Object ... | 2017-12-03 | Code |
| 129 | CornerNet511 (Hourglass-104, multi-scale) | 20.8 | No | CornerNet: Detecting Objects as Paired Keypoints | 2018-08-03 | Code |
| 130 | M2Det (ResNet-101, single-scale) | 20.5 | No | M2Det: A Single-Shot Object Detector based on Mu... | 2018-11-12 | Code |
| 131 | ExtremeNet (Hourglass-104, single-scale) | 20.4 | No | Bottom-up Object Detection by Grouping Extreme a... | 2019-01-23 | Code |
| 132 | Cascade R-CNN (ResNet-50-FPN+) | 20.3 | No | Cascade R-CNN: Delving into High Quality Object ... | 2017-12-03 | Code |
| 133 | Mask R-CNN (ResNet-101-FPN) | 20.1 | No | Mask R-CNN | 2017-03-20 | Code |
| 134 | DeformConv-R-FCN (Aligned-Inception-ResNet) | 19.4 | No | Deformable Convolutional Networks | 2017-03-17 | Code |
| 135 | SaccadeNet (DLA-34-DCN) | 19.2 | No | SaccadeNet: A Fast and Accurate Object Detector | 2020-03-26 | Code |
| 136 | Faster R-CNN (ImageNet+300M) | 17.5 | No | Revisiting Unreasonable Effectiveness of Data in... | 2017-07-10 | Code |
| 137 | CornerNet511 (Hourglass-52, single-scale) | 17 | No | CornerNet: Detecting Objects as Paired Keypoints | 2018-08-03 | Code |
| 138 | RefineDet512 (ResNet-101) | 16.6 | No | Single-Shot Refinement Neural Network for Object... | 2017-11-18 | Code |