YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

6 Jul 2022  ·  Chien-Yao Wang, Alexey Bochkovskiy, Hong-Yuan Mark Liao ·

YOLOv7 surpasses all known object detectors in both speed and accuracy in the range from 5 FPS to 160 FPS and has the highest accuracy 56.8% AP among all known real-time object detectors with 30 FPS or higher on GPU V100. YOLOv7-E6 object detector (56 FPS V100, 55.9% AP) outperforms both transformer-based detector SWIN-L Cascade-Mask R-CNN (9.2 FPS A100, 53.9% AP) by 509% in speed and 2% in accuracy, and convolutional-based detector ConvNeXt-XL Cascade-Mask R-CNN (8.6 FPS A100, 55.2% AP) by 551% in speed and 0.7% AP in accuracy, as well as YOLOv7 outperforms: YOLOR, YOLOX, Scaled-YOLOv4, YOLOv5, DETR, Deformable DETR, DINO-5scale-R50, ViT-Adapter-B and many other object detectors in speed and accuracy. Moreover, we train YOLOv7 only on MS COCO dataset from scratch without using any other datasets or pre-trained weights. Source code is released in https://github.com/WongKinYiu/yolov7.

PDF Abstract

Datasets


Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Real-Time Object Detection COCO YOLOv7-E6E (36 fps) FPS (V100, b=1) 36 # 18
box AP 56.8 # 1
Real-Time Object Detection COCO YOLOv7-D6 (44 fps) FPS (V100, b=1) 44 # 17
box AP 56.6 # 2
Real-Time Object Detection COCO YOLOv7-E6 (56 fps) FPS (V100, b=1) 56 # 13
box AP 56 # 3
Real-Time Object Detection COCO YOLOv7-W6 (84 fps) FPS (V100, b=1) 84 # 8
box AP 54.9 # 7
Real-Time Object Detection COCO YOLOv7-X (114 fps) FPS (V100, b=1) 114 # 6
box AP 53.1 # 14
Real-Time Object Detection COCO YOLOv7 (161 fps) FPS (V100, b=1) 161 # 4
box AP 51.4 # 22
Real-Time Object Detection COCO YOLOv7-tiny-SiLU (286 fps) FPS (V100, b=1) 286 # 1
box AP 38.7 # 39
Object Detection COCO test-dev YOLOv7 (161 fps) box AP 51.4 # 57
Object Detection COCO test-dev YOLOv7-X (114 fps) box AP 53.1 # 46
Object Detection COCO test-dev YOLOv7-W6 (84 fps) box AP 54.9 # 32
Object Detection COCO test-dev YOLOv7-E6 (56 fps) box AP 56 # 26
Object Detection COCO test-dev YOLOv7-D6 (44 fps) box AP 56.6 # 21
Object Detection COCO test-dev YOLOv7-E6E (36 fps) box AP 56.8 # 20
Object Detection COCO test-dev YOLOv7-tiny-SiLU (286 fps) box AP 38.7 # 179

Methods