tommyshelby4

Camera Parameters for the Monocular Video Demo

1

Where could I possibly find the camera parameters (intri.yml and extri.yml) for the monocular video demos demonstrated here https://chingswy.github.io/easymocap-public-doc/quickstart/quickstart.html? Thanks in advance.

Model Export

1

Hello everybody and congratulations on your amazing work. Could somebody confirm that **gen_wts_yoloV5.py** is replaced by **export_yolo_V5.py**?

Could somebody explicitly tell what the coordinate system of SMPL is? Forward, up and right vector?

When running the monocular demo, the output matrix containing the SMPL pose parameters has a dimension of 75, while in the multicamera setup (**mv1p**) the output vector dimension is 78....

Is it possible that during the postprocess I somehow get joint angles instead of 3D keypoints of SMPL joint positions?

3D keypoints in monocular demo

4

Is there a way I can obtain 3D keypoints instead of 2D keypoints in the monocular demo paradigm? Thanks in advance.

XYZ coordinate system

1

Is there any case that the 3D joint positions provided by the dataset are local transforms instead of global 3D positions?

Inference

2

If I want to use the model just for inference and not from training, is the dataset generation necessary?

Denoising mode representation

1

In the denoising mode, the input I get has dimension **(num_frames, 63)** which seems weird to me bearing in mind the PoseNDF uses quaternion representation as input. Shouldn't it be...