Object Detection on MS COCO (test-dev)
78.5mAP@.5DyHead+
Evaluation results
| Method | Links | |||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DyHead+Backbone=Swin-L, Iteration=2x, Extra data=true2021.06 | 78.5 | 60.6 | 43.9 | 64 | 74.2 | 66.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DyHeadBackbone=Swin-L, Iteration=2x, Extra data=false2021.06 | 77.1 | 58.7 | 41.7 | 62 | 72.8 | 64.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOv7-E6EParams=151.7M, FLOPs=843.2G, Size=1280, FPS (V100)=36, batch size=1, TensorRT=false, extra training data=false2022.07 | 74.4 | — | — | — | — | 62.1 | — | — | — | — | — | — | — | — | — | — | — | 56.8 | — | — | — | |
| YOLOv7-E6EParameters=151.7M, FLOPs=843.2G, Input Size=12802022.07 | 74.4 | — | 39.3 | 60.5 | 69 | 62.1 | — | — | — | — | — | — | — | — | — | — | 36 | 56.8 | — | — | — | |
| YOLOv7-E6EBackbone=E-ELAN [52], Input size=1280, FPS=362020.12 | 74.4 | 56.8 | 39.3 | 60.5 | 69 | 62.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientDet-D7xBackbone=EfficientNet-B7 [34], Size=1536, FPS=6.5*2020.11 | 74.3 | — | 37.2 | 57.9 | 68 | 59.9 | — | — | — | — | — | — | — | — | — | — | — | 55.1 | — | — | — | |
| ConvNeXt-XL (C-M-RCNN)FLOPs=1898.0G, Size=1280, FPS (A100)=8.6, batch size=1, TensorRT=false, extra training data=false2022.07 | 74.2 | — | — | — | — | 59.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOR-D6Params=151.7M, FLOPs=935.6G, Size=1280, FPS (V100)=34, batch size=1, TensorRT=false, extra training data=false2022.07 | 74.1 | — | — | — | — | 61.9 | — | — | — | — | — | — | — | — | — | — | — | 56.5 | — | — | — | |
| YOLOR-D6Parameters=151.7M, FLOPs=935.6G, Input Size=12802022.07 | 74.1 | — | 38.9 | 60.4 | 68.7 | 61.9 | — | — | — | — | — | — | — | — | — | — | 34 | 56.5 | — | — | — | |
| YOLOR-D6Backbone=CSPNet [47], Input size=1280, FPS=342020.12 | 74.1 | 56.5 | 38.9 | 60.4 | 68.7 | 61.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PRB-FPN6Backbone=E-ELAN [52], Input size=1280, FPS=312020.12 | 74.1 | 56.9 | 39 | 60.5 | 70 | 62.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CenterNet2+Backbone=Res2Net-101-DCN, Iteration=8x, Extra data=false2021.06 | 74 | 56.4 | 38.7 | 59.7 | 68.6 | 61.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOv7-D6Params=154.7M, FLOPs=806.8G, Size=1280, FPS (V100)=44, batch size=1, TensorRT=false, extra training data=false2022.07 | 74 | — | — | — | — | 61.8 | — | — | — | — | — | — | — | — | — | — | — | 56.6 | — | — | — | |
| YOLOv7-D6Parameters=154.7M, FLOPs=806.8G, Input Size=12802022.07 | 74 | — | 38.8 | 60.1 | 69.5 | 61.8 | — | — | — | — | — | — | — | — | — | — | 44 | 56.6 | — | — | — | |
| ConvNeXt-L (C-M-RCNN)FLOPs=1354.0G, Size=1280, FPS (A100)=10, batch size=1, TensorRT=false, extra training data=false2022.07 | 73.8 | — | — | — | — | 59.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MR-CenterNetBackbone=Swin-L, multi-scale testing=true, resolution_strategy=multi-resolution feature maps2022.04 | 73.7 | 57.1 | 38.7 | 59.2 | 71.3 | 62.4 | — | — | — | — | 40.9 | 67.4 | 72.2 | 54.8 | 75.1 | 86.8 | — | — | — | — | — | |
| YOLOv7-E6Params=97.2M, FLOPs=515.2G, Size=1280, FPS (V100)=56, batch size=1, TensorRT=false, extra training data=false2022.07 | 73.5 | — | — | — | — | 61.2 | — | — | — | — | — | — | — | — | — | — | — | 56 | — | — | — | |
| YOLOv7-E6Parameters=97.2M, FLOPs=515.2G, Input Size=12802022.07 | 73.5 | — | 38 | 59.9 | 68.4 | 61.2 | — | — | — | — | — | — | — | — | — | — | 56 | 56 | — | — | — | |
| YOLOv4-P7Backbone=CSP-P7, Size=1536, FPS=17/16*2020.11 | 73.4 | — | 38.4 | 59.4 | 67.7 | 60.8 | — | — | — | — | — | — | — | — | — | — | — | 55.5 | — | — | — | |
| YOLOR-E6Params=115.8M, FLOPs=683.2G, Size=1280, FPS (V100)=45, batch size=1, TensorRT=false, extra training data=false2022.07 | 73.4 | — | — | — | — | 61.1 | — | — | — | — | — | — | — | — | — | — | — | 55.8 | — | — | — | |
| YOLOR-E6Parameters=115.8M, FLOPs=683.2G, Input Size=12802022.07 | 73.4 | — | 38.4 | 59.7 | 67.7 | 61.1 | — | — | — | — | — | — | — | — | — | — | 45 | 55.8 | — | — | — | |
| YOLOv4-P7Backbone=CSP-P7, Epoch=450, multi-scale training=true, multi-scale testing=true2021.04 | 73.3 | 56 | 38.9 | 60 | 68.6 | 61.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOv4-P7Backbone=CSP-P7, multi-scale testing=true2022.04 | 73.3 | 56 | 38.9 | 60 | 68.6 | 61.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConvNeXt-B (C-M-RCNN)FLOPs=964.0G, Size=1280, FPS (A100)=11.5, batch size=1, TensorRT=false, extra training data=false2022.07 | 73.1 | — | — | — | — | 58.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VFNet-X-1200Backbone=R2-101-DCN, FPS=4.2, single-model single-scale=true2020.08 | 73 | 55.1 | 37.4 | 58.2 | 67 | 60.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| nmsModel=x6_12802021.12 | 72.8 | 55.1 | 37.8 | 58.9 | 66.5 | 60.4 | — | — | — | — | — | — | 71.5 | — | — | — | — | — | — | — | — | |
| CP-ClusterModel=x6_12802021.12 | 72.8 | 55.5 | 38.1 | 59.3 | 66.8 | 60.9 | — | — | — | — | — | — | 73.4 | — | — | — | — | — | — | — | — | |
| YOLOR-W6Params=79.8G, FLOPs=453.2G, Size=1280, FPS (V100)=66, batch size=1, TensorRT=false, extra training data=false2022.07 | 72.7 | — | — | — | — | 60.5 | — | — | — | — | — | — | — | — | — | — | — | 55.2 | — | — | — | |
| YOLOR-W6Parameters=79.8M, FLOPs=453.2G, Input Size=12802022.07 | 72.7 | — | 37.7 | 59.1 | 67.1 | 60.5 | — | — | — | — | — | — | — | — | — | — | 66 | 55.2 | — | — | — | |
| YOLOv4-P6Backbone=CSP-P6, Size=1280, FPS=32/30*2020.11 | 72.6 | — | 36.8 | 58.3 | 65.9 | 59.8 | — | — | — | — | — | — | — | — | — | — | — | 54.5 | — | — | — | |
| YOLOv7-W6Params=70.4M, FLOPs=360.0G, Size=1280, FPS (V100)=84, batch size=1, TensorRT=false, extra training data=false2022.07 | 72.6 | — | — | — | — | 60.1 | — | — | — | — | — | — | — | — | — | — | — | 54.9 | — | — | — | |
| YOLOv7-W6Parameters=70.4M, FLOPs=360.0G, Input Size=12802022.07 | 72.6 | — | 37.3 | 58.7 | 67.1 | 60.1 | — | — | — | — | — | — | — | — | — | — | 84 | 54.9 | — | — | — | |
| YOLOv4-P6Backbone=CSP-P6 [54], Input size=1280, FPS=322020.12 | 72.6 | 54.5 | 36.8 | 58.3 | 65.9 | 59.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientDet-D7Backbone=EfficientNet-B6 [34], Size=1536, FPS=8.2*2020.11 | 72.4 | — | 35.8 | 57 | 66.3 | 58.4 | — | — | — | — | — | — | — | — | — | — | — | 53.7 | — | — | — | |
| EfficientDet-D7xParams=77.0M, FLOPs=410.0G, Size=1536, FPS (V100)=6.5, batch size=1, TensorRT=false, extra training data=false2022.07 | 72.4 | — | — | — | — | 58.4 | — | — | — | — | — | — | — | — | — | — | — | 55.1 | — | — | — | |
| Swin-L (C-M-RCNN)FLOPs=1382.0G, Size=1280, FPS (A100)=9.2, batch size=1, TensorRT=false, extra training data=false2022.07 | 72.4 | — | — | — | — | 58.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DyHeadBackbone=ResNeXt-64x4d-101-DCN, Iteration=2x, multi-scale training=true, multi-scale testing=true2021.06 | 72.1 | 54 | 37.1 | 57.2 | 66.3 | 59.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade Mask R-CNNBackbone=Triple-ResNeXt152, Multi-scale testing=true2019.09 | 71.9 | 53.3 | 35.5 | 55.8 | 66.7 | 58.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TSDBackbone=SENet154-DCN, Multi-scale testing=true2020.10 | 71.9 | 51.2 | 33.8 | 54.8 | 64.2 | 56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Deformable DETRBackbone=ResNeXt-64x4d-101-DCN, Iteration=~4x, multi-scale training=true, multi-scale testing=true2021.06 | 71.9 | 52.3 | 34.4 | 54.4 | 65.6 | 58.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpineNetBackbone=SpineNet-190, Iteration=~40x, multi-scale training=true, multi-scale testing=false2021.06 | 71.8 | 52.1 | 35.4 | 55 | 63.6 | 56.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-B (C-M-RCNN)Params=145.0M, FLOPs=982.0G, Size=1280, FPS (A100)=10.7, batch size=1, TensorRT=false, extra training data=false2022.07 | 71.8 | — | — | — | — | 57.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VFNet-X-800Backbone=R2-101-DCN, FPS=8, single-model single-scale=true2020.08 | 71.6 | 53.7 | 34.4 | 57.5 | 67.5 | 58.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| nmsModel=l6_12802021.12 | 71.6 | 53.8 | 36.3 | 57.8 | 64.9 | 58.9 | — | — | — | — | — | — | 70.3 | — | — | — | — | — | — | — | — | |
| CP-ClusterModel=l6_12802021.12 | 71.6 | 54.1 | 36.6 | 58.1 | 65.3 | 59.4 | — | — | — | — | — | — | 72.4 | — | — | — | — | — | — | — | — | |
| EfficientDet-D6Backbone=EfficientNet-B6 [34], Size=1408, FPS=11*2020.11 | 71.5 | — | 34.9 | 56 | 65.4 | 57.2 | — | — | — | — | — | — | — | — | — | — | — | 52.6 | — | — | — | |
| EfficientDet-D7Backbone=B6, FPS=3.8, single-model single-scale=true2020.08 | 71.4 | 52.2 | 34.8 | 55.5 | 64.6 | 56.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientDetBackbone=Efficient-B7, Iteration=~50x, multi-scale training=true, multi-scale testing=false2021.06 | 71.4 | 52.2 | — | — | — | 56.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MR-CenterNetBackbone=Swin-L, multi-scale testing=false, resolution_strategy=multi-resolution feature maps2022.04 | 71.4 | 53.2 | 33.2 | 56.2 | 68.7 | 57.4 | — | — | — | — | 39.2 | 61.6 | 64 | 43.2 | 67.7 | 80.7 | — | — | — | — | — | |
| YOLOR-CSP-XParams=96.9M, FLOPs=226.8G, Size=640, FPS (V100)=87, batch size=1, TensorRT=false, extra training data=false2022.07 | 71.4 | — | — | — | — | 57.9 | — | — | — | — | — | — | — | — | — | — | — | 53 | — | — | — | |
| YOLOR-P6Params=37.2M, FLOPs=325.6G, Size=1280, FPS (V100)=76, batch size=1, TensorRT=false, extra training data=false2022.07 | 71.4 | — | — | — | — | 58.9 | — | — | — | — | — | — | — | — | — | — | — | 53.9 | — | — | — | |
| YOLOR-CSP-XParameters=96.9M, FLOPs=226.8G, Input Size=6402022.07 | 71.4 | — | 33.7 | 57.1 | 66.8 | 57.9 | — | — | — | — | — | — | — | — | — | — | 87 | 53 | — | — | — | |
| YOLOR-P6Parameters=37.2M, FLOPs=325.6G, Input Size=12802022.07 | 71.4 | — | 36.1 | 57.7 | 65.6 | 58.9 | — | — | — | — | — | — | — | — | — | — | 76 | 53.9 | — | — | — | |
| YOLOR-CSP-XBackbone=CSPNet [47], Input size=640, FPS=872020.12 | 71.4 | 53 | 33.7 | 57.1 | 66.8 | 57.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientDet-D6Backbone=B6, FPS=5.3, single-model single-scale=true2020.08 | 71.2 | 51.7 | 34.1 | 55.2 | 64.1 | 56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOv7-XParams=71.3M, FLOPs=189.9G, Size=640, FPS (V100)=114, batch size=1, TensorRT=false, extra training data=false2022.07 | 71.2 | — | — | — | — | 57.8 | — | — | — | — | — | — | — | — | — | — | — | 53.1 | — | — | — | |
| YOLOv7-XParameters=71.3M, FLOPs=189.9G, Input Size=6402022.07 | 71.2 | — | 33.8 | 57.1 | 67.4 | 57.8 | — | — | — | — | — | — | — | — | — | — | 114 | 53.1 | — | — | — | |
| YOLOv7-XBackbone=E-ELAN [52], Input size=640, FPS=1142020.12 | 71.2 | 53.1 | 33.8 | 57.1 | 67.4 | 57.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LSNet-CPVBackbone=R2-101-DCN, Epoch=24, multi-scale training=true, multi-scale testing=true, corner point verification=true2021.04 | 71.1 | 53.5 | 35.2 | 56.4 | 65.8 | 59.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LSNetBackbone=R2-101-DCN, multi-scale testing=true2022.04 | 71.1 | 53.5 | 35.2 | 56.4 | 65.8 | 59.2 | — | — | — | — | 39.8 | 66.2 | 71.4 | 53.7 | 74.1 | 85.3 | — | — | — | — | — | |
| PRB-FPN-MSPBackbone=MSPNet [55], Input size=640, FPS=942020.12 | 71.1 | 53.3 | 34.1 | 57.3 | 66.2 | 58.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GFLV2Backbone=R2-101-DCN, Epoch=24, multi-scale training=false, multi-scale testing=true2021.04 | 70.9 | 53.3 | 35.7 | 56.1 | 65.6 | 59.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GFLV2Backbone=R2-101-DCN, multi-scale testing=true2022.04 | 70.9 | 53.3 | 35.7 | 56.1 | 65.6 | 59.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MR-CenterNetBackbone=R2-101-DCN, multi-scale testing=true, resolution_strategy=multi-resolution feature maps2022.04 | 70.9 | 53.7 | 35.1 | 56 | 66.7 | 59.7 | — | — | — | — | 39.8 | 66.6 | 71.8 | 54.3 | 74.5 | 86.2 | — | — | — | — | — | |
| DyHeadBackbone=ResNeXt-64x4d-101-DCN, Iteration=2x, multi-scale training=true, multi-scale testing=false2021.06 | 70.7 | 52.3 | 35.1 | 56.2 | 63.4 | 57.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade Mask R-CNNBackbone=Dual-ResNeXt152, Multi-scale testing=true2019.09 | 70.6 | 52.8 | 34.9 | 55.4 | 65.3 | 58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EfficientDet-D5Backbone=EfficientNet-B5 [34], Size=1280, FPS=14*2020.11 | 70.5 | — | 33.9 | 54.7 | 64.1 | 56.7 | — | — | — | — | — | — | — | — | — | — | — | 51.5 | — | — | — | |
| RelationNet++Backbone=ResNeXt-64x4d-101-DCN, Multi-scale testing=true2020.10 | 70.4 | 52.7 | 35.8 | 55.3 | 64.7 | 58.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RetinaNetBackbone=S143 [6], Size=1280, FPS=102020.11 | 70.4 | — | 33.6 | 53.9 | 62.1 | 54.9 | — | — | — | — | — | — | — | — | — | — | — | 50.7 | — | — | — | |
| RelationNet++Backbone=ResNeXt-64x4d-101-DCN, Iteration=2x, multi-scale training=true, multi-scale testing=true2021.06 | 70.4 | 52.7 | 35.8 | 55.3 | 64.7 | 58.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PRB-FPN-ELANBackbone=ELAN [52], Input size=640, FPS=702020.12 | 70.4 | 52.5 | 33.4 | 56.2 | 65.8 | 57.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOv4-P5Backbone=CSP-P5, Size=896, FPS=43/41*2020.11 | 70.3 | — | 33.4 | 55.7 | 63.4 | 56.6 | — | — | — | — | — | — | — | — | — | — | — | 51.8 | — | — | — | |
| LSNet-CPVBackbone=R2-101-DCN, Epoch=24, multi-scale training=true, FPS=6.3, corner point verification=true, multi-scale testing=false2021.04 | 70.3 | 51.1 | 31.2 | 54.3 | 65 | 55.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RepPoints V2Backbone=X-101-64x4d-DCN, Epoch=24, multi-scale training=false, multi-scale testing=true2021.04 | 70.1 | 52.1 | 34.5 | 54.6 | 63.6 | 57.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RepPoints v2Backbone=ResNeXt-64x4d-101-DCN, Iteration=2x, multi-scale training=true, multi-scale testing=true2021.06 | 70.1 | 52.1 | 34.5 | 54.6 | 63.6 | 57.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RepPointsV2Backbone=X-101-64x4d-DCN, multi-scale testing=true2022.04 | 70.1 | 52.1 | 34.5 | 54.6 | 63.6 | 57.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PRB-FPN-CSPBackbone=CSP-P5 [54], Input size=640, FPS=1132020.12 | 70 | 51.8 | 32.6 | 55.5 | 64.6 | 56.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPYOLOE-XParams=98.4M, FLOPs=206.6G, Size=640, FPS (V100)=45, batch size=1, TensorRT=false, extra training data=false2022.07 | 69.9 | — | — | — | — | 56.5 | — | — | — | — | — | — | — | — | — | — | — | 52.2 | — | — | — | |
| PPYOLOE-XParameters=98.4M, FLOPs=206.6G, Input Size=6402022.07 | 69.9 | — | 33.3 | 56.3 | 66.4 | 56.5 | — | — | — | — | — | — | — | — | — | — | 45 | 52.2 | — | — | — | |
| SEPCBackbone=ResNeXt-64x4d-101-DCN, Iteration=2x, multi-scale training=true, multi-scale testing=false2021.06 | 69.8 | 50.1 | 31.3 | 53.3 | 63.7 | 54.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SNIP (model ensemble)Multi-scale testing=true2019.09 | 69.7 | 48.3 | 31.4 | 51.6 | 60.7 | 53.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TridentNetBackbone=ResNet-101-DCN2020.03 | 69.7 | 48.4 | 31.8 | 51.3 | 60.3 | 53.5 | — | — | — | — | — | — | — | — | — | — | 0.7 | — | — | — | — | |
| VFNetBackbone=R2-101-DCN, FPS=10.3, MSTrain image scale range=1333x[480:960], single-model single-scale=true2020.08 | 69.7 | 51.3 | 31.9 | 54.7 | 64.4 | 55.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PAABackbone=X-101-64x4d-DCN, Epoch=24, multi-scale training=false, multi-scale testing=true2021.04 | 69.7 | 51.4 | 34 | 53.8 | 64 | 57 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Deformable DETRBackbone=ResNeXt-64x4d-101-DCN, Iteration=~4x, multi-scale training=true, multi-scale testing=false2021.06 | 69.7 | 50.1 | 30.6 | 52.8 | 64.7 | 54.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TridentNetBackbone=R-101-DCN, multi-scale testing=true2022.04 | 69.7 | 48.4 | 31.8 | 51.3 | 60.3 | 53.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOv7Params=36.9M, FLOPs=104.7G, Size=640, FPS (V100)=161, batch size=1, TensorRT=false, extra training data=false2022.07 | 69.7 | — | — | — | — | 55.9 | — | — | — | — | — | — | — | — | — | — | — | 51.4 | — | — | — | |
| YOLOv7Parameters=36.9M, FLOPs=104.7G, Input Size=6402022.07 | 69.7 | — | 31.8 | 55.5 | 65 | 55.9 | — | — | — | — | — | — | — | — | — | — | 161 | 51.4 | — | — | — | |
| YOLOv7Backbone=E-ELAN [52], Input size=640, FPS=1612020.12 | 69.7 | 51.4 | 31.8 | 55.5 | 65 | 55.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| YOLOR-CSPParams=52.9M, FLOPs=120.4G, Size=640, FPS (V100)=106, batch size=1, TensorRT=false, extra training data=false2022.07 | 69.6 | — | — | — | — | 55.7 | — | — | — | — | — | — | — | — | — | — | — | 51.1 | — | — | — | |
| YOLOR-CSPParameters=52.9M, FLOPs=120.4G, Input Size=6402022.07 | 69.6 | — | 31.7 | 55.3 | 64.7 | 55.7 | — | — | — | — | — | — | — | — | — | — | 106 | 51.1 | — | — | — | |
| YOLOR-CSPBackbone=CSPNet [47], Input size=640, FPS=1062020.12 | 69.6 | 51.1 | 31.7 | 55.3 | 64.7 | 55.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PRB-FPNBackbone=CSPDarknet-53 [47], Input size=800x800, FPS=37.52020.12 | 69.5 | 48.9 | 30.8 | 55.9 | 60.2 | 55.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PRB-FPNBackbone=CSPDarknet-53, Input size=800x8002020.12 | 69.5 | 48.9 | 30.8 | 55.9 | 60.2 | 55.9 | — | — | — | — | — | — | — | — | — | — | 11.6 | — | — | — | — | |
| LSNet-CPVBackbone=X-101-64x4d-DCN, Epoch=24, multi-scale training=true, FPS=5.1, corner point verification=true, multi-scale testing=false2021.04 | 69.4 | 50.4 | 31 | 53.3 | 64 | 54.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VFNetBackbone=X-101-64x4d-DCN, FPS=6.7, MSTrain image scale range=1333x[480:960], single-model single-scale=true2020.08 | 69.3 | 50.8 | 31.6 | 54.2 | 64.4 | 55.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| nmsModel=m6_12802021.12 | 69.2 | 51.2 | 33.5 | 55.1 | 62.1 | 56.2 | — | — | — | — | — | — | 68.1 | — | — | — | — | — | — | — | — | |
| CP-ClusterModel=m6_12802021.12 | 69.2 | 51.5 | 33.7 | 55.4 | 62.5 | 56.7 | — | — | — | — | — | — | 70.2 | — | — | — | — | — | — | — | — | |
| MR-CenterNetBackbone=R2-101-DCN, multi-scale testing=false, resolution_strategy=multi-resolution feature maps2022.04 | 69.2 | 51.5 | 31 | 54.4 | 65.7 | 56.2 | — | — | — | — | 38.5 | 63.1 | 67.5 | 45.6 | 71.7 | 84.6 | — | — | — | — | — | |
| RelationNet++Backbone=ResNeXt-64x4d-101-DCN, Multi-scale testing=false2020.10 | 69 | 50.3 | 32.8 | 55 | 65.8 | 55 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |