Unique Head Count from a Monocular Video
Fig. 1: Influx and outflux for a ROI
Fig. 2: UIOC Results
UIOC refers to unique influx and out-flux count. UIOC is an algorithm to count uniquely many objects from a monocular video clip. UIOC counts one object only once as it enters and/or exits a region of interest (ROI). UIOC has excellent resistance to partial occlusions because it is not based on object tracking, which may be adversely affected by partial occlusions especially in a monocular video. UIOC applies short-term optical flow and Gaussian process regression. We achieved state-of-the-art counting accuracy on several benchmark videos (see publications for details). Our method applies to cell counts and other object counts too.
Related Publications
- S. Mukherjee, S. Gil and N. Ray, “Unique people count from monocular videos,” The Visual Computer, Vol.31, no.10, pp 1405-1417, October 2015.
- S. Mukherjee, N. Ray, S.T. Acton, ″Counting cells from microscopy videos without tracking individual cells,″ ISBI 2014.
- S. Mukherjee, "Unique people count," Doctoral Consortium at VISIGRAPP 2014. won First Prize.
- S. Mukherjee, B. Saha, I. Jamal, R. Leclerc, N. Ray, “A novel framework for automatic passenger counting,” Proceedings of IEEE ICIP 2011, pp. 2969-2972.
People Count from a Still Image
We designed a random projection-based descriptor to use with a classifier to estimate number of people present in an image. Our method achieved state-of-the-art accuracy (details of results in the publications).
Related Publications
- H. Foroughi, N. Ray, and H. Zhang, “Robust People Counting using Sparse Representation and Random Projection,” Pattern Recognition, Special Issue on Discriminative Feature Learning from Big Data for Visual Recognition, Pattern Recognition, vol.48, no.10, pp.3038-3052, 2015.
- H. Foroughi, N. Ray and H. Zhang, “People counting with image retrieval using compressed sensing,” IEEE ICASSP 2014.