search

leoxiaobin / deep-high-resolution-net.pytorch

The project is an official implementation of our CVPR2019 paper "Deep High-Resolution Representation Learning for Human Pose Estimation"
https://jingdongwang2017.github.io/Projects/HRNet/PoseEstimation.html
MIT License
4.29k stars 908 forks source link

mpii submission #12

Open yellowcheng opened 5 years ago

yellowcheng commented 5 years ago

hi, i use test.json to creat a [7247. 16. 2] pred.mat, can i submitt pred.mat to mpii official website to have a test result?

leoxiaobin commented 5 years ago

You should convert to MPII's format following his instruction at http://human-pose.mpi-inf.mpg.de/#evaluation for submission.

aliYaHf commented 5 years ago

@yellowcheng Hi！ Are you using the official data of the MPII dataset, or the processed data provided by Xiao, how is the training result?

yellowcheng commented 5 years ago

@aliYaHf yes, and the result on mpii val is the same as the paper, but i still can't figure out how to prepare my test result and submit to the mpii website

738654805 commented 5 years ago

@yellowcheng Have you solve this problem? I meet the same problem.

U-C-J commented 4 years ago

Same problem. In test.json provided by HRnet, the test.json size is 7247. But in https://github.com/anibali/eval-mpii-pose, the test.h5 size is 11731. Also, the center and scale are different in test.json and test.h5 in that post. Now I am really confused. Did anyone figure it out? Thank you.

Github3G commented 3 years ago

Same problem. In test.json provided by HRnet, the test.json size is 7247. But in https://github.com/anibali/eval-mpii-pose, the test.h5 size is 11731. Also, the center and scale are different in test.json and test.h5 in that post. Now I am really confused. Did anyone figure it out? Thank you.

Do you have resolved this problem? I meet same proble, and the center and scale are differnet, but i use the h5 to test for submitt. at the other questiion https://github.com/leoxiaobin/deep-high-resolution-net.pytorch/issues/51, one person think "center" values are annotated manually

MaxChu719 commented 3 years ago

Same problem. In test.json provided by HRnet, the test.json size is 7247. But in https://github.com/anibali/eval-mpii-pose, the test.h5 size is 11731. Also, the center and scale are different in test.json and test.h5 in that post. Now I am really confused. Did anyone figure it out? Thank you.

Do you have resolved this problem? I meet same proble, and the center and scale are differnet, but i use the h5 to test for submitt. at the other questiion https://github.com/leoxiaobin/deep-high-resolution-net.pytorch/issues/51, one person think "center" values are annotated manually

I used the test.h5 to generate the 11731 test results and the PCKh score i got is only 89.9 instead of 92.3 reported in the paper. I have implemented the multi-scale test on the valid set and I got pckh 90.75 so the multi-scale output should be fine.

STRUGGLE1999 commented 1 year ago

hi, i use test.json to creat a [7247. 16. 2] pred.mat, can i submitt pred.mat to mpii official website to have a test result?

hi, i use test.json to creat a [7247. 16. 2] pred.mat, can i submitt pred.mat to mpii official website to have a test result?

Hello, Could you tell me how to convert pred.mat to json file for visualision ?

STRUGGLE1999 commented 1 year ago

You should convert to MPII's format following his instruction at http://human-pose.mpi-inf.mpg.de/#evaluation for submission.

Hello, Could you tell me how to get a test result for mpii, I get a pred.mat, but couldn't visualize it.