Abstract:
Deep networks estimate people densities in images for modern crowd counting methods. Few video sequences use temporal consistency, and those that do only apply weak smoothness constraints across frames. This paper proposes estimating people flows across image locations between consecutive images and inferring people densities from them instead of directly regressing them. This lets us impose stronger population conservation constraints. Thus, it improves performance without requiring a more complex architecture. It also lets us use people flow and optical flow correlations to improve results. We demonstrate that leveraging people conservation constraints in both space and time allows active learning of a deep crowd counting model with fewer annotations. This significantly reduces annotation cost while achieving similar performance to full supervision.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
