Text-to-face retrieval using self-supervised learning

Description :

The goal of this project is to retrieve images containing a given person based on text queries. Given a dataset of images with captions, we want to learn a mapping between person names and face images. Our dataset is unlabelled, which means that we don’t have access to this mapping. The proposed approach is to learn such a mapping with self-supervised learning. In this project, the candidate will work with face recognition models based on convolutional neural networks (such as FaceNet, SphereFace or InsightFace), and deep neural networks based word embeddings.

Deliverables: codebase, web app (optional)

The project requires programming skills in Python, especially the PyTorch library.

The candidate should have experience with deep learning models, or keen to learn about them.

Pro-active in learning and trying new things, everyday.

Send me your CV: [email protected].