search

lightas / ICCV19_Pose_Guided_Occluded_Person_ReID

This is the pytorch implementation and dataset of the ICCV2019 paper "Pose-Guided Feature Alignment for Occluded Person Re-identification"
146 stars 29 forks source link

The pose information of occluded dukemtmc-reid training data #7

Open kimx3966 opened 4 years ago

kimx3966 commented 4 years ago

Hi, first thank you for sharing your code. I'm using official Alphapose v0.2.0 version code which I assume that you are using in this project. However, most of results are very similar to yours, but I have slightly different result from occluded dukemtmc-reid dataset. I have the pose information of test data including gallery and query, but I cannot find one for training data. So, can you provide the pose information of occluded dukemtmc-reid training data? And, can you provide weights (duc_se.pth and yolov3-spp.weights) used in this project? Please upload your files to Google Drive if it is possible. Thank you in advance.

lightas commented 4 years ago

Hi, please refer to these two links: https://drive.google.com/file/d/1D47msNOOiJKvPOXlnpyzdKA3k6E97NTC/view?usp=sharing

https://drive.google.com/file/d/1OPORTWB2cwd5YTVBX-NE8fsauZJWsrtW/view?usp=sharing

kimx3966 commented 4 years ago

Thanks for sharing the weight. Although I run your alphapose code with the weight, I still got slightly different result. Can you provide the pose data of training data? Thank you in advance for your help.