The author introduces a novel approach, Stealing Stable Diffusion (SSD), for robust monocular depth estimation by leveraging stable diffusion and self-training mechanisms. The core thesis is to enhance depth estimation in challenging conditions using generative diffusion models.
This work presents Depth Anything, a highly practical solution for robust monocular depth estimation. By scaling up the dataset with large-scale unlabeled data and employing effective training strategies, the model exhibits impressive generalization ability across extensive unseen scenes.
A novel robust depth estimation framework, D4RD, that incorporates a customized contrastive learning scheme for diffusion models to mitigate performance degradation in complex environments.
Developing depth estimation models that can maintain satisfactory performance under real-world corruptions and perturbations, such as adverse weather conditions, sensor failure, and noise contamination.