Event Recognition in Unconstrained Video using Multi-Scale Deep Spatial Features

Event recognition in an unconstrained video is a challenging problem due to its complex nature in the field of computer vision. In this paper, we have designed a new technique using deep learning framework to recognize an event during video classification. This technique is enough capable for modeling multiscale spatial information to correctly classify events either in short or long videos. Three Convolutional Neural Networks (CNN) followed by Long Short Term Memory (LSTM) architectures are used to extract multi-scale spatial features from each video. The main contribution of this work is the design and development of deep learning framework that can model unconstrained videos based on several important aspects. The performance of the proposed system is tested on popular and challenging benchmark database namely, Columbia Consumer Videos (CCV). The extensive experiment shows that the proposed deep learning framework is able to recognize an event in the video quite successfully compared to many state-of-the art methods.