[NeurIPS 2025] Tracking and Understanding Object Transformations If you like our project, please give us a star ⭐ on GitHub for the latest update. 💡 Description Dataset Visualizations: GitHub Paper: arXiv:2511.04678 Project Page: tubelet graph.github.io Project Repository: GitHub Point of Contact: Yihong Sun 📊 Dataset Overview VOST TAS (TrackAnyState) is an extended version of the VOST validation set with explicit transformation annotations for tracking and understanding object state changes in videos. This dataset enables evaluation of video understanding systems on their ability to track objects through physical transformations and describe resulting state changes. The dataset contains: 57 video instances 108 transformations 293 annotated resulting objects Each video sequence is annotated with temporal segments corresponding to object state transformations, including action descriptions and segmentation masks for resulting objects. 📁 Structure This dataset contains annotated videos and images for object transformation tracking with detailed state change annotations. The directory structure and file descriptions are as follows: Annotations/ : Contains segmentation masks for the…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy