Seatle, 17-21/06/2024.
The IEEE/CVF Computer Vision and Pattern Recognition Conference (CVPR) is among the most prestigious annual conferences for computer vision with more than 10.000 attendees and several co-located workshops.
Scientists from Fraunhofer HHI, partner of iToBoS, attended this year’s CVPR conference in Seattle and presented papers that resulted from research conducted throughout the iToBoS project.
This includes the papers titled “Reactive Model Correction: Mitigating Harm to Task-Relevant Features via Conditional Bias Suppression” and “Understanding the (Extra-)Ordinary: Validating Deep Model Decisions with Prototypical Concept-based Explanations”, which were both presented at the workshop “Safe Artificial Intelligence for All Domains (SAIAD)”.
Whereas the first paper introduces a reactive model correction approach, the latter presents a novel global XAI approach. Moreover, the paper “PURE: Turning Polysemantic Neurons Into Pure Features by Identifying Relevant Circuits” was presented at the 3rd Explainable AI for Computer Vision (XAI4CV) Workshop, introducing a method to disentangle polysemantic neurons.
Find out more at https://cvpr.thecvf.com/Conferences/2024.