davidelanz/vanishing_points

•Updated almost 6 years ago

Vanishing Point Detection using Inverse Gnomonic Projection [Caffe on Python 2.7]

Image

269

davidelanz/vanishing_points repository overview

⁠Deep Learning for Vanishing Point Detection Using an Inverse Gnomonic Projection

⁠Copyright disclaimer

The copyright in this software is being made available under the BSD LICENSE⁠ Copyright (c) 2017-present Leibniz University Hannover (LUH) Institut fuer Informationsverarbeitung (TNT)

The code provided here was adapted from:
@inproceedings{kluger2017deep,
  title={Deep learning for vanishing point detection using an inverse gnomonic projection},
  author={Kluger, Florian and Ackermann, Hanno and Yang, Michael Ying and Rosenhahn, Bodo},
  booktitle={German Conference on Pattern Recognition (GCPR)},
  year={2017}
}
The paper can be found on arXiv⁠. Refer to fkluger/vanishing_points_2017⁠ for additional information and resources.

⁠Getting Started

⁠Requirements

Caffe 1.0-RC5⁠ on Python 2.7
ImageMagick
Python requirements.txt packages

⁠Setup

Launch the docker container from https://hub.docker.com/repository/docker/davidelanz/vanishing_points⁠
```
docker pull davidelanz/vanishing_points
```

Test the software:

cd /home/vanishing_points
pytest test.py

⁠Manual Setup

Launch a docker container from https://hub.docker.com/r/bvlc/caffe⁠ (Python 2.7 + Caffe)
```
docker pull bvlc/caffe:cpu
```
or any other suitable tag, e.g.:
```
docker pull bvlc/caffe:gpu
```

Initialize the repository

git clone https://github.com/Davidelanz/vanishing_points.git
cd vanishing_points
pip install -r requirements.txt

Download the CNN weights and image mean files from the releases⁠ and put them into the cnn folder.
Adjust config.py so that it contains the path to your Caffe installation and the paths where you store the benchmark datasets⁠.

⁠Run

You can run the vanishing point detector on the example images and visualise the results. Computation may take a few moments. Adjust the gpu_id in the main.py file if necessary. Then run: