My research interests are:
- Machine Learning
- Computer Vision
- Information Retrieval
- Programming languages and compilers
- GPU programming
I am currently working on classification and detection of steel defect using machine learning and machine vision techniques. The work is done in collaboration with ArcelorMittal and Centre for Mathematical Morphology of ParisTech.
I make an extensive use of convolutional-based neural networks in my works and I am interested in their extensions to narrow the gap between computer vision and machine learning. To extensively test this systems I use GPU implementations which run orders of magnitude faster than the corresponding fairly optimized CPU versions using arrayfire as a foundation framework.
I am very interested also in the metric learning problem, especially in the setting of similarity-sensitive hashing. To tackle this problems I proposed an extension of the siamese framework (DrLim) to produce binary representations and also a coupled version of the framework to address the multimodal case.