Thank you for your work. I want to use the predicted handpose information as the pseudo label of my model. I have two questions.
Assume that the predicted pose params, shape params, 3d joint and 2D joint of image1 are P_1, S_1, 3D_1, 2D_1; Image_1 is transformed, that is, transforms. Resize (224), to get image_2.
Can P_1, S_1, 3D_1, and 2D_1 be used as the label of image_2?
Does the predicted 3d joint and 2D joint represent the coordinate position of the hand in SMPL space and image sapce? Xyz coordinate?
@BetterZH
joints_smplcoord are joints in origin SMPL space.
joints_imgcoord are joints in image space. You can take the first two dim for pixel location.
Thank you for your work. I want to use the predicted handpose information as the pseudo label of my model. I have two questions.