search

facebookresearch / InterWild

Official PyTorch implementation of "Bringing Inputs to Shared Domains for 3D Interacting Hands Recovery in the Wild", CVPR 2023
Other
159 stars 16 forks source link

How to train MSCOCO for comparison models. #12

Open redorangeyellowy opened 11 months ago

redorangeyellowy commented 11 months ago

Hello. Thank you for your wonderful research.

In Table 5 of the main paper, you compared the result of training on not only InterHand2.6M but also MSCOCO for three comparison models (IHMR, Zhang et al, IntagHand).

However, MSCOCO seems to have no camera parameters unlike InterHand2.6M, so I wonder how you were able to train MSCOCO for the above three comparison models.

mks0601 commented 11 months ago

3D pseudo-GTs from NeuralAnnot (one of my prev. works) provides MANO parameters along with camera parameters.

NeuralAnnot: https://github.com/mks0601/NeuralAnnot_RELEASE Getting MANO from MSCOCO: https://github.com/facebookresearch/InterWild/blob/9d0e6ca4ea6bb301d3a2c9c3ff201683617db79b/data/MSCOCO/MSCOCO.py#L174