Fine-grained Classification and Segmentation for Fashion Images

The detection, segmentation, and classification of clothes in fashion images is a well-addressed issue on deep learning research. The challenge is to achieve similar results using images taken from street and in-store sites. For such images, the variety of human positions and high diversity of clothes features decreases the results compared to the former tests. This project aims to investigate which image features and deep learning models can better execute garment detection, segmentation, and classification for in-store fashion images. The partner organization will have the opportunity to work with high qualified human resource with competence to execute high technology research. Additionally, the organization will contribute to the training of a worker in a real-life company environment thus with the capacity to act in the technology market.

Intern: 
José Renato Villela Dantas
Superviseur universitaire: 
Laurent Charlin
Province: 
Quebec
Université: 
Partner University: 
Discipline: 
Programme: