If you are interested, code and weights are already available!!
Paper: arxiv.org/pdf/2503.08344 📜
Code: github.com/lmur98/DIV_F... 💻

Moreover, the distillation of video-language features allows to capture detailed affordances, where previous approaches failed due to they used CLIP features

Our model enables detailed segmentation and maintains consistent understanding over time. For instance, we can track dynamic objects along the video

We propose to decompose the egocentric scene into persistent, dynamic, and actor-based components while integrating both image and video-language features.

Egocentric videos are characterized by dynamic interactions and a strong dependence on the wearer engagement with the environment. Traditional approaches often focus on isolated clips or fail to integrate rich semantic and geometric information, limiting scene comprehension.