1. grad-cam 설치
!pip install grad-cam2. 파일 추가
https://github.com/open-mmlab/mmdetection/pull/7987
[WIP] Support Grad Free and Grad Based CAM by hhaAndroid · Pull Request #7987 · open-mmlab/mmdetection
We support cam visualization in mmdet. Currently, I have tested RetinaNet, Faster RCNN, Mask RCNN and YOLOX. Currently grad_free and grad_based methods are supported as shown below: GRAD_FREE_METHO...
github.com
demo/vis_cam.py
mmdet/utils/det_cam_visualizer.py
3. assert 삭제
demo/vis_cam.py
assert args.no_norm_in_bbox is False, 'If not norm in bbox, the ' \'visualization result ' \'may not be reasonable.'
4. python 실행
!python /content/mmdetection/demo/vis_cam.py /content/drive/MyDrive/12.jpg /content/mmdetection/configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py /content/drive/MyDrive/epoch_24.pth --no-norm-in-bbox --method gradcam --out-dir /content/sample_data'딥러닝' 카테고리의 다른 글
| Intra-class distance (0) | 2022.10.25 |
|---|---|
| [COCO] 데이터 확인 (1) | 2022.10.04 |
| [MMdetection] AP for each class (0) | 2022.09.21 |
| [MMdetection] samples_per_gpu (0) | 2022.09.16 |
| convert_to_coco (0) | 2022.09.14 |