Learning to Generalize with Self-Supervision

Add to your list(s) Download to your calendar using vCal

• Tatiana Tommasi, Politecnico di Torino & Italian Institute of Technology
• Friday 04 December 2020, 10:00-11:00
• https://bham-ac-uk.zoom.us/j/89097300074?pwd=ZnVZTmF3bGNYeWpVSFRkSEZpRnJhQT09.

If you have a question about this talk, please contact Martin Rudorfer.

Although deep networks have significantly increased the performance of visual recognition methods, it is still challenging to achieve a good level of robustness across visual domains, which is crucial for real-world applications. In many practical tasks collecting annotated samples may be very costly, but at the same time using models trained from data belonging to a different domain will produce only poor results. To tackle this issue, research on Domain Adaptation (DA) and Generalization (DG) has flourished over the last decade with several approaches based on feature alignment, generative and adversarial solutions. In this talk I will present a new point of view on the DA and DG settings that considers self-supervision as an auxiliary powerful tool to adapt and generalize across domains. Specifically the talk will show how solving a jigsaw puzzle or recognizing the orientation of an image can improve robustness and support generalization of models learned on photos, cartoons or sketches. We will also see how this beneficial effect extends from object recognition to detection and to 3d object part segmentation. Moreover the proposed strategy holds even in the most challenging open-world scenario with only a partial overlap between the class sets of source and target domains.

This talk is part of the IRLab Seminars: Robotics, Computer Vision & AI series.

This talk is included in these lists:

• IRLab Seminars: Robotics, Computer Vision & AI
• https://bham-ac-uk.zoom.us/j/89097300074?pwd=ZnVZTmF3bGNYeWpVSFRkSEZpRnJhQT09

Note that ex-directory lists are not shown.