mmdetection常用命令

（注意：文中--eval bbox适用于目标检测模型，其他模型--eval 需要换参数）

1. Test

python tools/test.py configs/ballon/mask_rcnn_r50_caffe_fpn_mstrain-poly_1x_balloon.py work_dirs/mask_rcnn_r50_caffe_fpn_mstrain-poly_1x_balloon.py/latest.pth --eval bbox

2. Train

python tools/train.py configs/ballon/mask_rcnn_r50_caffe_fpn_mstrain-poly_1x_balloon.py

3. Loss Analysis

• Plot the classification loss of some run.

  python tools/analyze_logs.py plot_curve log.json --keys loss_cls --legend loss_cls

• Plot the classification and regression loss of some run, and save the figure to a pdf.

• python tools/analyze_logs.py plot_curve log.json --keys loss_cls loss_bbox --out losses.pdf

• Compare the bbox mAP of two runs in the same figure.

  python tools/analyze_logs.py plot_curve log1.json log2.json --keys bbox_mAP --legend run1 run2

• Compute the average training speed.

  python tools/analyze_logs.py cal_train_time log.json [--include-outliers]

4. Output json result file 

python tools/test.py [configs_file] [pth] --eval-options "jsonfile_prefix=results_name" --eval bbox

This command will output "results_name.bbox.json" file.

此格式文件中含有image_id, bbox, score 和 category_id 四类信息。

5. Output pkl file

python tools/test.py [configs_file] [pth] --out results_name.pkl --eval bbox

This command will output "results_name.pkl" file.

6. Output metric

python tools/metric.py [configs_file] [pkl_file] 

7. coco_error_analysis

tools/coco_error_analysis.py analyzes COCO results per category and by different criterion. It can also make a plot to provide useful information.

python tools/coco_error_analysis.py [coco_result_json_file] [output_dir] --ann [ann_json_file] 

From: https://mmdetection.readthedocs.io/en/latest/useful_tools.html#error-analysis
Eval matrix 指标介绍: https://cocodataset.org/#detection-eval
 

https://github.com/open-mmlab/mmdetection/blob/master/docs/2_new_data_model.md

https://github.com/open-mmlab/mmdetection/blob/master/docs/useful_tools.md

版权声明：本文为CSDN博主「zxrzhang」的原创文章，遵循CC 4.0 BY-SA版权协议，转载请附上原文出处链接及本声明。
原文链接：https://blog.csdn.net/xin_rong/article/details/112055616