About the Project
Recognition of human actions from video has many applications including: enhancing the well-being, security, designing better social spaces.
The extraction of optical ﬂow and its further statistical and syntactical processing provides very important clues towards human interaction recognition. This research proposal will extend classical approaches in action recognition towards identifying human interactions between two or groups of people. Deep Learning through Convolutional Neural Networks (CNN) has proved successful in many computer vision applications including for video processing.
During the ﬁrst stage the PhD study will adapt using transfer learning existing deep Convolutional Networks used for optical ﬂow processing for recognition human interactions.
During the second stage, a new recursive CNN will be developed by combining networks specialised in optical ﬂow processing, moving region segmentation and human activity recognition,aiming to identify interactions between participants in the given scene.
The data used for the research undertaken during this PhD study will be provided by existing databases, by video cameras (CCTV for example) or by data provided by mobile video cameras, such as those mounted on drones.
 K. Stephens, A. G. Bors, "Group Activity Recognition on Outdoor Scenes," Proc. of 13th IEEE International Conference on Advanced Video and Signal-based Surveillance (AVSS), Colorado Springs, CO, USA, August 2016, pp. 59-65
 A. G. Bors and I. Pitas, "Optical Flow Estimation and Moving Object Segmentation Based on Median Radial Basis Function Network,"IEEE Trans. on Image Processing, vol. 7, no. 5, pp. 693-702, May 1998.
 K. Simonyan, A. Zisserman, “Two-Stream Convolutional Networks for Action Recognition in Videos,”Advances in Neural Information Processing Systems, 2014
Based on your current searches we recommend the following search filters.
Based on your current search criteria we thought you might be interested in these.