search

princeton-vl / pose-hg-train

Training and experimentation code used for "Stacked Hourglass Networks for Human Pose Estimation"
Other
575 stars 185 forks source link

Evaluation on MPII Human Pose Dataset #18

Closed Litchiware closed 7 years ago

Litchiware commented 7 years ago

Hi, @anewell, I want to use the evaluation script evaluatePCKh.m downloaded on MPII Human Pose Dataset to measure my model's performance, but I could not find the annolist_dataset_v12.mat file loaded in this line of the script.

load([p.gtDir '/annolist_dataset_v12'],'annolist');

Do you know how to get this file, thanks.

anewell commented 7 years ago

You can get that file from the MPII website from the download page you linked - "Annotations (12.5 MB)"

anewell commented 7 years ago

Oh sorry, I wasn't paying attention, that's not my code actually, so not sure what that file is referring to.

You can also use the python code in this repository under src/pypose to do evaluation. (Edit: actually maybe not a great idea to use that code at the moment, it hasn't been updated to work with recent changes, instead check out the evaluation code in the other repository pose-hg-demo)

Litchiware commented 7 years ago

Thanks for your reply. I think I can use pose-hg-train/src/misc/predictMPII.m to generate pred_keypoints_mpii.mat, and send this file to MPII dataset team and get evaluation results from them, right?

anewell commented 7 years ago

Yep, you can do that. The file 'predictMPII.m' needs to be updated a little bit, just a small modification to read in the predictions properly. I will push that in a little bit.

In general, it is good to make sure that you can also do evaluation on validation predictions since you don't want to submit predictions on the official MPII test set too often.

Litchiware commented 7 years ago

Thanks a lot, I think this issue can be closed.