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
710 stars 130 forks source link

dataset question #56

Open asw91666 opened 3 years ago

asw91666 commented 3 years ago

Hi! Could i know why didn't you use dataset MPII, UP-3D additionally?

mks0601 commented 3 years ago

Both are small scale, and UP-3D does not provide good 3D fits. For me, just using MSCOCO was enough.

asw91666 commented 3 years ago

Did you do the test with those datasets in I2L-meshnet?

mks0601 commented 3 years ago

I can't get your question. All dataset descriptions are included in the paper.