Revealing Occlusions with 4D Neural Fields
This addresses the challenge of handling occlusions in video understanding tasks for computer vision systems, representing an incremental improvement with a novel method for a known bottleneck.
The paper tackles the problem of object permanence in dynamic scenes by introducing a framework that learns 4D visual representations from monocular RGB-D to persist objects through occlusions, showing successful occlusion resolution on two released datasets without architectural changes.
For computer vision systems to operate in dynamic situations, they need to be able to represent and reason about object permanence. We introduce a framework for learning to estimate 4D visual representations from monocular RGB-D, which is able to persist objects, even once they become obstructed by occlusions. Unlike traditional video representations, we encode point clouds into a continuous representation, which permits the model to attend across the spatiotemporal context to resolve occlusions. On two large video datasets that we release along with this paper, our experiments show that the representation is able to successfully reveal occlusions for several tasks, without any architectural changes. Visualizations show that the attention mechanism automatically learns to follow occluded objects. Since our approach can be trained end-to-end and is easily adaptable, we believe it will be useful for handling occlusions in many video understanding tasks. Data, code, and models are available at https://occlusions.cs.columbia.edu/.