/Deep-Learning-Traffic_Sign_Recognizer

First project to understand Deep learning

Primary LanguagePython

Deep-Learning-Traffic_Sign_Recognizer

MLGroup16

The repository for group 16 for the course Machine Learning. Detecting traffic signs on pictures

Used Libraries

XLSXWriter

numpy, pylab, skimage, sklearn, scipy,

OpenCV

For shape matching, you must install OpenCV. I used this to get it working:

Gohlke maintains Windows binaries for many Python packages, including OpenCV 3.0 with Python 3.x bindings! See here:

http://www.lfd.uci.edu/~gohlke/pythonlibs/#opencv

To install, just download the 64-bit or 32-bit .whl file appropriate for your system, then run pip install [filename]. Then the instruction import cv2 should work in your Python 3.x interpreter.

Mahotas

For calculating the zernike moments, we used a library called Mahotas. You can download a build of it from the website:

http://www.lfd.uci.edu/~gohlke/pythonlibs/#mahotas

To install, just download the 64-bit or 32-bit .whl file appropriate for your system, then run pip install [filename]. Then the instruction import cv2 should work in your Python 3.x interpreter.

Lasagne

Lasagne is a lightweight library to build and train neural networks in Theano. How it can be installed is explained in their docs:

http://lasagne.readthedocs.org/en/latest/user/installation.html

IDE: Pycharm (JetBrains)

How to run

All variables can be declared in main.py and feature extractors can be selected.

tsr.make_submission(train_images_path=train_images_dir, test_images_path=test_images_dir, output_file_path="test.xlsx", feature_extractors=feature_extractors, size=64)

print(tsr.local_test(train_images_path=train_images_dir, feature_extractors=feature_extractors, k=2, nr_data_augments=1, size=64))