/Color-Recognition-CNN

CNN for color recognition of cars

Primary LanguagePythonMIT LicenseMIT

Color-Recognition-CNN

There 9 classes( ['black', 'blue', 'brown', 'green', 'pink', 'red', 'silver', 'white', 'yellow']).

  1. For the training you need to download my dataset https://drive.google.com/file/d/1rD235inKHQxC0tVUOlNZuIfOXc8PXoaq/view?usp=sharing and just run train.py script (Don't forget to change num of classes and other parameters if you use your own dataset)
  2. After training you can use test.py script. It's cropping car's from images with using object detection from Tensorflow Object Detection API (Also you need to download one of the models from Tensorflow Object Detection API for test.py script, for example: http://download.tensorflow.org/models/object_detection/faster_rcnn_inception_v2_coco_2018_01_28.tar.gz)

Original paper: https://arxiv.org/pdf/1510.07391.pdf