OccJEPA: Occlusion as Masking for Amodal Video Segmentation
Amodal video segmentation asks for the complete mask of an object in every frame, including the parts hidden behind other objects.
ProofPaper ↗
Key points
- The best current methods generate these masks with fine-tuned video diffusion models, which need many denoising steps and an extra depth network, and give a different answer for every sample.
- Self-supervised video world models of the JEPA family are trained on a closely related task: predicting the representation of masked space-time regions from the visible ones.
- We therefore treat occlusion as masking.
- The resulting model, OccJEPA (Occlusion-as-masking JEPA), is deterministic and 33.5x faster than the diffusion-based state of the art, and more accurate on SAIL-VOS, on TAO-Amodal, including a held-out test split (+4.00 AP50) and visible masks tracked by SAM2, and on Kubric.
Sources (1)
- [1]OccJEPA: Occlusion as Masking for Amodal Video SegmentationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 01:40 AM
Amodal video segmentation asks for the complete mask of an object in every frame, including the parts hidden behind other objects.
The best current methods generate these masks with fine-tuned video diffusion models, which need many denoising steps and an extra depth network, and give a different answer for every sample.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 20, 2026Exploring Temporal Context-Aware Physical Grounding in Latent World Models
- Oct 20, 2026MoDDiT: Simple Token Routing in Diffusion Transformers
- Oct 20, 2026[AINews] DeepSeek DeepInfer: Open Source SOTA Inference running 155T toks/day/cluster making $22m/day at ~80% margin for a 81 day payback
- Oct 20, 2026Do Diffusion LLMs Scale? Testing Mercury 3 + Decide
- Oct 20, 2026Slot-JEPA: Object-Structured Prediction Interfaces for Latent World Models
- Oct 19, 2026Representing short-term movement to reveal animal identity and learning progress