deep-high-resolution-net.pytorch icon indicating copy to clipboard operation
deep-high-resolution-net.pytorch copied to clipboard

Published 20 hours ago • verified publisher icon leoxiaobin

How to test the model on COCO test-dev set?

Open touhou-ayaya opened this issue 4 years ago • 3 comments

How to test the model on COCO test-dev set? 请问您是如何在COCO test-dev数据集上进行测试？

touhou-ayaya avatar Dec 20 '20 06:12 touhou-ayaya

I also want to test on test-dev, how do you implement it？

Whj-cv avatar May 17 '21 07:05 Whj-cv

I also want to test on test-dev, how do you implement it？

For simple, you can download the detector results GoogleDrive first.

Then, modify the config file (namely, .yaml), where TEST_SET is set to 'test-dev', COCO_BBOX_FILE is set to the path of COCO_test-dev2017_detections_AP_H_609_person.json, last but important, USE_GT_BBOX should be switched to 'false'.

You can obtain a test result on COCO test-dev by running the following code python tools/test.py --cfg experiments/coco/hrnet/w32_256x192_adam_lr1e-3.yaml TEST.MODEL_FILE models/pytorch/pose_coco/pose_hrnet_w32_256x192.pth The result is a json file, and you can follow the procedure to obtain the evaluation score.

touhou-ayaya avatar May 19 '21 02:05 touhou-ayaya

I also want to test on test-dev, how do you implement it？

For simple, you can download the detector results GoogleDrive first.

Then, modify the config file (namely, .yaml), where TEST_SET is set to 'test-dev', COCO_BBOX_FILE is set to the path of COCO_test-dev2017_detections_AP_H_609_person.json, last but important, USE_GT_BBOX should be switched to 'false'.

You can obtain a test result on COCO test-dev by running the following code python tools/test.py --cfg experiments/coco/hrnet/w32_256x192_adam_lr1e-3.yaml TEST.MODEL_FILE models/pytorch/pose_coco/pose_hrnet_w32_256x192.pth The result is a json file, and you can follow the procedure to obtain the evaluation score.

thanks，i will try to achieve it

Whj-cv avatar May 19 '21 02:05 Whj-cv