Are training set and val set splitted correctly? about simple-effective-3dpose-baseline

[lck1201]

Hi, please refer to paper A simple yet effective baseline for 3d human pose estimation， I follow his evaluation

[lucas946]

Sorry, it's my fault, your setting is right. However, Human 3.6M dataset contains two type of 3d coordinates, which are D3_Positions_mono and D3_Positions_mono_universal and they are both in camera coordinate system. My question is, which type of 3d coordinates do you use?

[lck1201]

Hi, the dataset I use is the internal processed version, so I don't meet your problem. But I think you could back-project this two types of points to 2D plane to see which one should be used

[lucas946]

Thanks a lot, I will figure it out myself.