Person re-identification through face detection from videos using Deep Learning
Vimala Mathew, Tom Toby, Anu Mary Chacko, Arasu Udhayakumar · 2019
In a networked video surveillance system, the videos collected from different surveillance cameras are stored in a centralized server. These videos can be made available for scrutiny, during an event of security threat. Real time analysis of these surveillance videos can be helpful in preventing crimes in the areas under surveillance. A major challenge in video analytics is object detection from video frames. Identifying people in the videos has many applications like handling security through surveillance cameras, crime detection, personalized assistance for the needy, product purchase promotion, employee monitoring, etc. Accurate person reidentification from videos has huge potential that can revolutionize the way businesses work today. This paper discusses a method for person re-identification from the videos collected from surveillance cameras. The critical tasks in this process are face detection from videos and prediction of persons using Convolutional Neural Network models developed using the cropped face images from the face detection stage.