Overview / Usage

Person Re-identification involves detection if same person exists in a given pair of images.
One of key problems is that the images can be captured from various viewpoints leading to a very low accuracy.
Other problems are partial occlusions, illumination variation in the pair of images.

I tend to discover performance of deep CNNs on this problem using normalized cross correlation to combine the features extracted from the image.

Methodology / Approach

Deep nets

Collaborators