如何从训练中打印出yolov7中最高的mAP

Question

当我使用以下命令训练yolov7时

!python train.py --data {dataset.location}/data.yaml --weights 'yolov7_training.pt' --device 0 --hyp "hyp.scratch.custom.yaml"

输出：

Epoch   gpu_mem       box       obj       cls     total    labels  img_size
100/299     10.5G   0.05475   0.02964         0   0.08439       243       640: 100% 15/15 [00:11<00:00,  1.31it/s]
Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 1/1 [00:00<00:00,  3.27it/s]
all          15         178       0.465       0.607       0.509       0.155
Epoch   gpu_mem       box       obj       cls     total    labels  img_size
101/299     10.5G   0.05034   0.03036         0    0.0807       249       640: 100% 15/15 [00:11<00:00,  1.30it/s]
Class      Images      Labels           P           R      mAP@.5  mAP@.5:.95: 100% 1/1 [00:00<00:00,  4.01it/s]
all          15         178       0.509       0.584       0.489       0.152

我使用进行检测

!python detect.py --weights runs/train/exp2/weights/best.pt --conf 0.9 --source {dataset.location}/valid/images

如何打印训练中的最高mAP@.5？

Answer 1

我设置了wanb，并添加了一个带有以下行的Weave以获得

runs.map((row) => row.history["metrics/mAP_0.5(B)"].max)