Depth Anywhere: Enhancing 360 Monocular Depth Estimation via Perspective Distillation and Unlabeled Data Augmentation
This work addresses a domain-specific challenge in virtual reality and autonomous navigation by enhancing 360-degree depth estimation, though it is incremental as it builds on existing perspective methods.
The paper tackles the problem of depth estimation in 360-degree imagery, which suffers from a lack of labeled data, by proposing a framework that uses perspective depth models as teachers to generate pseudo labels via cube projection, resulting in significant accuracy improvements on benchmark datasets like Matterport3D and Stanford2D3D.
Accurately estimating depth in 360-degree imagery is crucial for virtual reality, autonomous navigation, and immersive media applications. Existing depth estimation methods designed for perspective-view imagery fail when applied to 360-degree images due to different camera projections and distortions, whereas 360-degree methods perform inferior due to the lack of labeled data pairs. We propose a new depth estimation framework that utilizes unlabeled 360-degree data effectively. Our approach uses state-of-the-art perspective depth estimation models as teacher models to generate pseudo labels through a six-face cube projection technique, enabling efficient labeling of depth in 360-degree images. This method leverages the increasing availability of large datasets. Our approach includes two main stages: offline mask generation for invalid regions and an online semi-supervised joint training regime. We tested our approach on benchmark datasets such as Matterport3D and Stanford2D3D, showing significant improvements in depth estimation accuracy, particularly in zero-shot scenarios. Our proposed training pipeline can enhance any 360 monocular depth estimator and demonstrates effective knowledge transfer across different camera projections and data types. See our project page for results: https://albert100121.github.io/Depth-Anywhere/