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

Benchmark Model Rank Results
2d-object-detection-on-ceymoYOLOv7#2mAP: 69.5
object-detection-on-cocoYOLOv7-D6 (44 fps)#39box mAP: 56.6
object-detection-on-cocoYOLOv7-E6 (56 fps)#45box mAP: 56
object-detection-on-cocoYOLOv7-W6 (84 fps)#50box mAP: 54.9
object-detection-on-cocoYOLOv7-X (114 fps)#66box mAP: 53.1
object-detection-on-cocoYOLOv7 (161 fps)#79box mAP: 51.4
object-detection-on-coco-oYOLOv7-E6E#15Average mAP: 32.0Effective Robustness: 6.42
pedestrian-detection-on-dvtodYOLOv7 (Thermal)#7mAP: 77.8
pedestrian-detection-on-dvtodYOLOv7 (Visible)#8mAP: 35.3
real-time-object-detection-on-cocoYOLOv7-E6E(1280)#5box AP: 56.8FPS (V100, b=1): 36
real-time-object-detection-on-cocoYOLOv7-D6(1280)#6box AP: 56.6FPS (V100, b=1): 44
real-time-object-detection-on-cocoYOLOv7-E6(1280)#9box AP: 56FPS (V100, b=1): 56
real-time-object-detection-on-cocoYOLOv7-W6(1280)#15box AP: 54.9FPS (V100, b=1): 84
real-time-object-detection-on-cocoYOLOv7-X#29box AP: 53.1FPS (V100, b=1): 114