About Train pose2mesh freihand

hongsukchoi / Pose2Mesh_RELEASE

Official Pytorch implementation of "Pose2Mesh: Graph Convolutional Network for 3D Human Pose and Mesh Recovery from a 2D Human Pose", ECCV 2020

MIT License

666 stars 69 forks source link

Open lihaonan1120 opened 1 year ago

lihaonan1120 commented 1 year ago

Hello Team,

When I tried to reproduce this article, I found that my loss value had a jump in about 16 epoch.

Current epoch 15, lr: 0.001， Epoch15 Loss: 0.0290； Current epoch 16, lr: 0.001， Epoch16 Loss: 0.0500。

May I ask if my reproduction result is the same as your original result？Has the training been over-fitted?

hongsukchoi commented 1 year ago

Hi,

I don't remember the details... I guess it's not a big problem, if the final epoch's performance is the same with our paper.