Crowd Emotion Analysis Using 2D ConvNets

Timely detection of the crowd emotions can lead to an effective crowd management measure. Crowd can exhibit multiple behaviour having separate emotions underneath it. Accurate emotion analysis of the crowd from a crowded image plays a critical role in the prudent examination of the crowded area. A popular non-intrusive solution for crowd behavior analysis is based on videos obtained through ambient camera, and the corresponding methods usually require a large dataset to train a classifier and are inclined to be influenced by the image quality. Detection of emotion of crowds becomes more pertinent in case of religious and political rallies. Peaceful conduct of these events is important for saving of human lives. The paper presents a novel method of deciphering crowd emotions using 2D convolutional neural network (ConvNets). This framework is then used to predict the crowd emotions, which leads to the subtle hints about the overall behaviour of crowd. The paper presents a suitable classifier for crowd emotion using the self-curated datasets for proving the concept that emotions can be extracted using ConvNets. Experiments have verified the proposed scheme on crowd behavior benchmark with fair accuracy.