Awesome

Form2Fit

Code for the paper

Form2Fit: Learning Shape Priors for Generalizable Assembly from Disassembly<br/> Kevin Zakka, Andy Zeng, Johnny Lee, Shuran Song<br/> arxiv.org/abs/1910.13675<br/> ICRA 2020

This repository contains:

• The Form2Fit Benchmark
  • Code to download and process the benchmark datasets.
  • Code to evaluate any model's performance on the benchmark test set.
• Code to reproduce the paper results:
  • Architectures, dataloaders and losses for suction, place and matching networks.
  • Planner module for intergrating all the outputs.
  • Baseline implementation.

If you find this code useful, consider citing our work:

@inproceedings{zakka2020form2fit,
  title={Form2Fit: Learning Shape Priors for Generalizable Assembly from Disassembly},
  author={Zakka, Kevin and Zeng, Andy and Lee, Johnny and Song, Shuran},
  booktitle={Proceedings of the IEEE International Conference on Robotics and Automation},
  year={2020}
}

Documentation

• setup
• about the Form2Fit benchmark
• reproducing paper results
• evaluating a trained model
• model weights
• conventions

Todos

• Add processed generalization partition (combinations, mixtures and unseen) to benchmark.
• Add code for training the different networks.

Note

This is not an officially supported Google product.