/deep-learning-latent-space

Learning Latent Space of Compliant Objects with Deep Learning for Robotic Grasping

Primary LanguagePythonMIT LicenseMIT

deep-learning-latent-space

Learning Latent Space of Compliant Objects with Deep Learning for Robotic Grasping

Installation

Make sure you have CUDA before proceeding with the installation.

conda env create -f torch-env.yml
conda activate torch

Download data

Download the shapenet and shapenet_car folders from this link, and save them to the folder source/data.

Build chamfer distance

conda activate torch
cd source/utils/pcd/chamfer
python setup.py install

Usage

  • Navigate to the /dl_models folder: cd source
  • Activate the conda environment: conda activate torch
  • Start visdom for visualization during training: visdom
  • [Optional] See training summaries on tensorboardX: tensorboard --logdir experiments/<name_of_experiment>
  • Train the network: python main.py configs/pcn.json