Effect of image colourspace on performance of convolution neural networks

Recently the term Deep Learning has been creating a lot of interest in the fields of Artificial Intelligence, Computer Vision and Natural Language Processing. And especially the Convolution Neural Networks (CNN) are giving state of art results in image recognition, scene understanding, object detection and image description etc. Generally in CNN the processing of images is done in RGB colourspace even though we have many other colourspaces available. In this paper we try to understand the effect of image colourspace on the performance of CNN models in recognizing the objects present in the image. We evaluate this on CIFAR10 dataset, by converting all the original RGB images into four other colourspaces like HLS, HSV, LUV, YUV etc. To compare results we have trained AlexNet with fixed set of parameters on all five colourspaces, including RGB. We have observed that LUV colourspace is the best alternative to RGB colourspace to use with CNN models with almost equal performance on the test set of CIFAR10 dataset. While YUV colourspace is the worst to use with CNN models.