In this project we have explored the use of neural networks to solve two common computer vision tasks using the PyTorch framework. We have been given a video-surveillance dataset containing images of multiple persons each of which is captured multiple times by different cameras along with a set of annotations that specify attributes of each person such as age, gender and clothing. The first part of the project consists in building a multi-class classifier to predict such attributes for each image. In the second part of the project, we are asked to solve a person re-identification problem where a query image of a person is given and all the images of the same person must be retrieved from a collection of images.
For a more in depth explanation of procedure and results see the report.