https://github.com/kevinzakka/form2fit

[ICRA 2020] Train generalizable policies for kit assembly with self-supervised dense correspondence learning.
https://github.com/kevinzakka/form2fit

deep-learning perception pytorch robotics self-supervised-learning

Last synced: 19 days ago
JSON representation

[ICRA 2020] Train generalizable policies for kit assembly with self-supervised dense correspondence learning.

• Host: GitHub
• URL: https://github.com/kevinzakka/form2fit
• Owner: kevinzakka
• License: mit
• Created: 2019-09-23T05:28:53.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2020-10-12T09:17:38.000Z (about 4 years ago)
• Last Synced: 2024-04-14T18:18:23.092Z (7 months ago)
• Topics: deep-learning, perception, pytorch, robotics, self-supervised-learning
• Language: Python
• Homepage: https://form2fit.github.io/
• Size: 53.5 MB
• Stars: 81
• Watchers: 13
• Forks: 23
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Form2Fit

Code for the paper

**[Form2Fit: Learning Shape Priors for Generalizable Assembly from Disassembly][1]**


*Kevin Zakka, Andy Zeng, Johnny Lee, Shuran Song*


[arxiv.org/abs/1910.13675][2]


ICRA 2020







This repository contains:

- The [Form2Fit Benchmark](docs/about_benchmark.md)

  - Code to [download and process](#data) the benchmark datasets.

  - Code to [evaluate](docs/evaluate_benchmark.md) any model's performance on the benchmark test set.

- Code to [reproduce](docs/paper_code.md) 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](docs/setup.md)

- [about the Form2Fit benchmark](docs/about_benchmark.md)

- [reproducing paper results](docs/paper_code.md)

- [evaluating a trained model](docs/evaluate_benchmark.md)

- [model weights](docs/model_weights.md)

- [conventions](docs/conventions.md)

### 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.

[1]: https://form2fit.github.io/

[2]: https://arxiv.org/abs/1910.13675