markpp / PoseFromPointClouds

Tensorflow and Pytorch implementations for pose predictions from point clouds.

PoseFromPointClouds

Tensorflow and Pytorch implementations for pose predictions from point clouds. The Tensorflow implementation is a simplified version of the one used for producing the results for "Cutting Pose Prediction from Point Clouds". The pytorch implementation is used for ongoing work on GNNs.

Usage

1. edit train.py to select between datasets and network_type (optional)
2. python3 train.py
3. python3 test.py
4. python3 evaluate.py

TODO

1. write more instructions
2. Add visualization scripts

Context

The models are part of the systems shown here.