Computer Vision – ECCV2024 18th European Conference

Federico Cocchi and Marcella Cornia (University of Modena and Reggio Emilia) participated in the 18th European Conference on Computer Vision (ECCV 2024), presenting the paper “Safe-CLIP: Removing NSFW Concepts from Vision-and-Language Models.”

The study addresses the challenge of enhancing the safety of vision-and-language models, such as CLIP, by reducing their sensitivity to NSFW (not safe for work) inputs. The proposed approach “unlearns” unsafe concepts by fine-tuning the model on synthetically generated safe and unsafe data, improving its reliability in sensitive and trustworthy contexts. The research demonstrates how this method can be effectively integrated with pre-trained generative models for safer applications.