https://github.com/rpng/calc2.0

CALC2.0: Combining Appearance, Semantic and Geometric Information for Robust and Efficient Visual Loop Closure
https://github.com/rpng/calc2.0

deep-learning slam variational-autoencoder

Last synced: 5 months ago
JSON representation

CALC2.0: Combining Appearance, Semantic and Geometric Information for Robust and Efficient Visual Loop Closure

• Host: GitHub
• URL: https://github.com/rpng/calc2.0
• Owner: rpng
• License: apache-2.0
• Created: 2019-03-04T20:28:48.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2022-11-21T21:02:08.000Z (almost 3 years ago)
• Last Synced: 2023-03-08T23:13:30.248Z (over 2 years ago)
• Topics: deep-learning, slam, variational-autoencoder
• Language: Python
• Homepage:
• Size: 30.1 MB
• Stars: 84
• Watchers: 16
• Forks: 17
• Open Issues: 9
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
## CALC2.0

Convolutional Autoencoder for Loop Closure 2.0.

To get started, download the COCO dataset and the "stuff" annotations, then run `dataset/gen_tfrecords.py`.

Make sure to unzip the tar in the dataset directory first.

Doing this will generate the sharded tfrecord files as well as `loss_weights.txt`.

After that you can train with `calc2.py`.

Check the --mode options in calc2.py to see what else you can do, like PR curves and finding the best model in a directory.

If you use this code for your research, please cite our paper:

```

@InProceedings{Merrill2019IROS,

  Title                    = {{CALC2.0}: Combining Appearance, Semantic and Geometric Information for Robust and Efficient Visual Loop Closure},

  Author                   = {Nathaniel Merrill and Guoquan Huang},

  Booktitle                = {2019 International Conference on Intelligent Robots and Systems (IROS)},

  Year                     = {2019},

  Address                  = {Macau, China},

  Month                    = nov,

}

```