search

mks0601 / I2L-MeshNet_RELEASE

Official PyTorch implementation of "I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image", ECCV 2020
MIT License
720 stars 128 forks source link

Handling KeyError exceptions of smpl_params #34

Closed andgitchang closed 3 years ago

andgitchang commented 3 years ago

Hi, I found there are indeed some keyerror exceptions of smpl_params as you handled. I'd like to realize how it will be processed when we call the function get_smpl_coord, since smpl_params are None. Is it feasible for me to manually populate zeros onto smpl_param? Or just ignore this none smpl_param annotation? Thanks.

mks0601 commented 3 years ago

Hi, you can see that when smpl_param is None, all variables related to smpl_param are set to zero tensors (see here). During the training, smpl_joint_trunc, smpl_mesh_trunc, is_valid_fit will make the loss from smpl_param zero.

andgitchang commented 3 years ago

I apologize I didn't trace the code thoroughly... Thanks for your explanation.