This project implements various different deep learning based interest point methods. Right now only superpoint.
- Python >= 3.5
- PyTorch >= 0.4
- tqdm (Optional for
test.py
) - tensorboard >= 1.7.0 (Optional for TensorboardX)
- tensorboardX >= 1.2 (Optional for TensorboardX)
- opencv-python
- parmap
- json_minify
For overall project structure see the README in source.
Please modify the configs/magic_point.json. You should set the data_dir as that sets where the generated synthetic shapes are created. Optionally modify the save_dir(where checkpoint files go) and log dir (where log files for tensorboard go).
Then run
python train.py --config configs/magic_point.json
If using this please cite the originally SuperPoint paper
@inproceedings{detone18superpoint,
author = {Daniel DeTone and
Tomasz Malisiewicz and
Andrew Rabinovich},
title = {SuperPoint: Self-Supervised Interest Point Detection and Description},
booktitle = {CVPR Deep Learning for Visual SLAM Workshop},
year = {2018},
url = {http://arxiv.org/abs/1712.07629}
}
The Superpoint portions of this project was ported from https://github.com/rpautrat/SuperPoint