Abstract

Video classification contains lot of challenges compared to the text data, due to its special quality known as temporal and spatial features. It is one of the processes of video analytics and creates more interest among researchers. The interestingness is rapidly penetrated based on the real-time demand of the applications. Video classification process requires large memory and high computation demand are the major complexities. In this case, we use tiny video representation to compress the temporal dimension of the video. Further, it is well suitable for the scenery and activities recognition scenarios. In the original work of tiny videos author used the Red Green Blue (RGB) colour descriptor. Generally, the colour descriptor can play a significant role on the image analysis, whereas it shows the poor performance on videos. In this paper we have examined the performance of Histogram of Oriented Gradients (HOG), Histogram of Optical Flow (HOF) descriptors with existing RGB colour descriptor. The proposed tiny video representation along with HOG and HOF descriptors are showing the better performance than colour descriptor.

Paper Details