SteeredMarigold: Steering Diffusion Towards Depth Completion of Largely Incomplete Depth Maps
This addresses a practical problem for robotics and AR/VR applications by enabling accurate depth completion in real-world scenarios with missing data, representing a novel approach but incremental in the broader field of depth completion.
The paper tackles depth completion for largely incomplete depth maps from RGB-D sensors, introducing SteeredMarigold, a training-free, zero-shot method that uses sparse depth points to steer a diffusion model, achieving state-of-the-art performance on the NYUv2 dataset with remarkable robustness against incompleteness.
Even if the depth maps captured by RGB-D sensors deployed in real environments are often characterized by large areas missing valid depth measurements, the vast majority of depth completion methods still assumes depth values covering all areas of the scene. To address this limitation, we introduce SteeredMarigold, a training-free, zero-shot depth completion method capable of producing metric dense depth, even for largely incomplete depth maps. SteeredMarigold achieves this by using the available sparse depth points as conditions to steer a denoising diffusion probabilistic model. Our method outperforms relevant top-performing methods on the NYUv2 dataset, in tests where no depth was provided for a large area, achieving state-of-art performance and exhibiting remarkable robustness against depth map incompleteness. Our source code is publicly available at https://steeredmarigold.github.io.