search

JinluZhang1126 / MixSTE

Official implementation of CVPR 2022 paper(MixSTE: Seq2seq Mixed Spatio-Temporal Encoder for 3D Human Pose Estimation in Video)
192 stars 16 forks source link

Processed Data of HRNet #22

Open flyyyyer opened 1 year ago

flyyyyer commented 1 year ago

Could you provide the processed data of HRNet? The VideoPose only contains CPN data.

nies14 commented 1 year ago

@flyyyyer you can check PoseAug to get the processed data of HRNet

flyyyyer commented 1 year ago

The setting of PoseAug is different from this work, they use 16 joints rather than 17 joints.

qiqiApink commented 1 year ago

@flyyyyer I have the same question. Did you solve this?

kyang-06 commented 1 year ago

Hi all, you can use fine-tuned HRNet at here: https://github.com/Nicholasli1995/EvoSkeleton As I have evaluated, it achieves better results than CPN.

wenwen12321 commented 1 year ago

Hi all, you can use fine-tuned HRNet at here: https://github.com/Nicholasli1995/EvoSkeleton As I have evaluated, it achieves better results than CPN.

Hi @kyang-06 ,

Could you please point me out how to use the processed data of HRNet (twoDPose_HRN_train.npy) (which is provided by https://github.com/Nicholasli1995/EvoSkeleton) to train in this project?

Because the pre-processing between these two projects seen to be a little different?

Thank you so much!!

Looking forward to your reply.