Chao Wen

Hi, @jin-s13 . I have signed the CLA. I will continuously submit some commits in this branch recently to improve the PR.

Code will be significantly refactored to fit v1.x, so it will take some extra time.

It would be nice to have Depth-Based 3D Hand Pose Estimation methods like [A2J](https://github.com/zhangboshen/A2J).

Hi @TowerTowerLee, thanks for your interest. 1. `y_train` contains pointcloud coordinates xyz and normal vector, so `point` and `normal` are represented by `y_train`. 2. Since our code does not have...

Hi everyone. Since the dataset we use is also from the author of 3DR2N2, the original data was rendered using blender. The camera parameters for rendering are just like @KnightOfTheMoonlight...

Transferred to human mesh generation may require a better initial mesh and more vertices points. In theory, it is feasible, while the implementation may be designed more specific for the...

Hi, @ttaa9 . The sampling steps are done using Meshlab script, so they're not shown in the code.

I think the problem may be related to wrong encoding format. Please load .dat files using pickle. If you are using python3 to read .dat file, you can add encoding...

All those files are used during data preparation, the number indicate the order of steps. We first (1) sample the points, then (2) calculate the normal, then (3) perform a...

Yes, `.dat` files dump using pickle from `.xyz` files generated by `generate_data.py`. You can read in the `.dat` file and using pickle to load it, you will find that it...