single mention · not yet filed
Masked depth modeling for RGB-D depth completion
Masked Depth Modeling is a masked image modeling pretraining method that predicts depth maps instead of reconstructing appearances, operating on RGB-D inputs within an encoder–decoder framework.1 It embeds RGB and depth separately into spatially aligned patch tokens so the encoder jointly learns appearance context and geometric cues from depth.2
How this definition was written
Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.
masked image modeling [9] within an encoder–decoder framework, but shifts the learning objective from appearance reconstruction to depth map prediction by operating on RGB-D inputs
— Masked Depth Modeling for Spatial PerceptionWe apply separate patch embedding layers to the two input modalities
— Masked Depth Modeling for Spatial Perception
Also called Masked Depth Modeling (MDM).