mmdetection 使用测试时值过低

最新推荐文章于 2024-10-08 15:59:01 发布

chao_xy

最新推荐文章于 2024-10-08 15:59:01 发布

阅读量229

点赞数 5

文章标签：深度学习人工智能

本文链接：https://blog.csdn.net/chao_xy/article/details/141929679

版权

运行 ./tools/dist_test.sh work_dirs/faster-rcnn_r50_fpn_1x_coco/faster-rcnn_r50_fpn_1x_coco.py work_dirs/faster-rcnn_r50_fpn_1x_coco/epoch_150.pth 1 --out work_dirs/faster-rcnn_r50_fpn_1x_coco/faster-rcnn_r50_fpn_1x_coco.pkl

奇怪的bug

+-----------+-------+--------+--------+-------+-------+-------+
| category | mAP   | mAP_50 | mAP_75 | mAP_s | mAP_m | mAP_l |
+-----------+-------+--------+--------+-------+-------+-------+
| 1 | 0.0   | 0.001 | 0.0    | nan   | 0.0   | 0.0   |
| 2 | 0.001 | 0.01   | 0.0    | 0.008 | 0.0   | 0.002 |
| 3 | 0.0   | 0.0    | 0.0    | nan   | nan   | 0.0   |
| 4     | 0.002 | 0.006 | 0.0    | nan   | nan   | 0.002 |
| 5 | 0.0   | 0.0    | 0.0    | nan   | nan   | 0.0   |
+-----------+-------+--------+--------+-------+-------+-------+

需要将config文件（faster-rcnn_r50_fpn_1x_coco.py）中从num_classes=5（你的种类数）改为num_classes=80

再运行就是正常的