AUVSI ADLC Training Samples Creation.

This package creates target samples for training the ADLC algortihm. The targets are created accoording to the AUVSI SUAS competition. The sofware was developed and used by the TAS team during the 2015 and 2016 competitions.

Installation

Linux

> git clone https://github.com/amitibo/auvsi-targets.git
> cd auvsi_targets
> sudo apt-get install python-opencv virtualenvwrapper ttf-dejavu
> mkvirtual --system-site-packages auvsi # we need access to opencv from the virtualenv
(auvsi) > pip install -r requirements.txt
(auvsi) > ./install_aggdraw

Windows

Python - Recommended to install using a distibution like Anaconda.
opencv - Install from Unofficial python binaries. Tested with version >= 2.4.10 and 3.1.0
aggdraw - (used for the image processing project): Install from Unofficial python binaries.

To install do:

> git clone https://github.com/amitibo/auvsi-targets.git
> cd auvsi_targets
> pip install -r requirements.txt

How to Use

To create shape trainig samples run ./create_patches.py
To create letter training samples run ./create_letter_samples.py

Input images are under DATA/resized_images and DATA/renamed_images, flight data is under DATA/flight_data.

The shape samples and letter samples are created under DATA/train_images and DATA/train_letter respectively. Each sample is made of an image file and a corresponding label file.

License