Code for "Neural Body: Implicit Neural Representations with Structured Latent Codes for Novel View Synthesis of Dynamic Humans" CVPR 2021 best paper candidate
I actually converted SMPL to COCO today, that was for 3d.
I have another question.
Usually for conversion of 3d to 2d, we use the Camera Projection Matrix. Now after Human reconstruction using Neural Body, if I want to extract and convert the 3d joints into 2d, which camera matrix should I use?
The original camera using which we generated the SMPL. ( Eg. For zjumocap, we have 4 cameras - 1, 7, 13, 11)?
OR
We should use the "rendering camera" aka the Virtual Camera that we use for reconstructing the Human Figure?
Which one? If you can provide some more insights on this, it would be wonderful.
Kindly do let me know.
Hello @pengsida,
I actually converted SMPL to COCO today, that was for 3d.
I have another question.
Usually for conversion of 3d to 2d, we use the Camera Projection Matrix. Now after Human reconstruction using Neural Body, if I want to extract and convert the 3d joints into 2d, which camera matrix should I use?
Which one? If you can provide some more insights on this, it would be wonderful. Kindly do let me know.