visualize training result for mmdetection

mmdetection_visualize_v1

It's a very simple version for visualizing the training result produced by mmdetection.

Readme

The program supports drawing six training result and the most important evaluation tool:PR curve(only for VOC now)

1. loss_rpn_bbox

2. loss_rpn_cls

3. loss_bbox

4. loss_cls

5. loss

6. acc

7. PR_curve

8. F-measure

Installation

1. Clone it
   git clone https://github.com/Stephenfang51/mmdetection_visualize

There will be total 5 files(json directory, output directory, visualize.py, mean_ap_visualize.py, voc_eval_visualize.py)

• put voc_eval_visualize.py under /mmdetection/tools/

• put mean_ap_visualize.py under mmdetection/mmdet/core/evaluation/

How to use

six training result

1. After training finished, you will have work_dir directory in your mmdetection directory
2. take the latest json file and put into json directory in mmditection_visualize directory
3. command python visualize.py json/xxxxxxxlog.json in terminal
4. check the output directory, Done !

PR curve and F-measure

1. make sure voc_eval_visualize.py and mean_ap_visualize.py settled down
2. command as usual like python tools/voc_eval_visualize.py {your pkl file} {your network configs file}
   • example python tools/voc_eval_visualize.py result.pkl ./configs/faster_rcnn_r101_fpn_1x.py
3. check the /mmdetection main directory, you will see the PR_curve_each_class.png there， Done !

GitHub