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

How to only predict the hand like the teaser? #70

Closed buaacyw closed 3 years ago

buaacyw commented 3 years ago

Thanks for your great work! image Could you please tell me how to predict the hand like this teaser? image I run demo.py on the image and get the above result. This is a (640,480) image and I have changed the box to (0,0,320,360). All the model follows your installation instruction and run your input image well. Thanks!

mks0601 commented 3 years ago

demo code is only for the body. You should change it to use MANO.

mks0601 commented 3 years ago

https://github.com/mks0601/I2L-MeshNet_RELEASE#freihand-dataset

buaacyw commented 3 years ago

Thanks for your quick reply!

buaacyw commented 3 years ago

https://github.com/mks0601/I2L-MeshNet_RELEASE/blob/5d495593fa99e3e44af0289964a7da7284fd9876/demo/demo.py#L124 I'm not sure but do you mean that change this visualize function and the pre-trained model? And why is mano needed? Function vis_img seems to be able to visualize hand mesh. @mks0601