images_lr and smpl_param

I am currently working with the MVHuman dataset, specifically the images_lr frames and the corresponding smpl_param files. While exploring the data, I noticed that the number of SMPL parameter files does not match the number of frames in images_lr. For example, in subject 22236222, there are hundreds of images_lr frames, but only a subset of frames have corresponding SMPL .pkl files.

Could you please clarify the reason for this discrepancy? Is it due to SMPL fitting failures on certain frames, or is there another intended design choice in the dataset? Understanding this will help me properly align the frames and parameters for my experiments.

Compare the frame names and counts under images_lr with the available smpl_param .pkl files for subject 22236222. Check the dataset documentation or generation notes first; done means documenting whether missing parameters reflect fitting failures or an intentional subset and how users should align the data.

Written by the indexing model from the issue text.