A Low Memory Footprint Quantized Neural Network for Depth Completion of Very Sparse Time-of-Flight Depth Maps.
Xiaowen JiangValerio CambareriGianluca AgrestiCynthia Ifeyinwa UgwuAdriano SimonettoFabien CardinauxPietro ZanuttighPublished in: CoRR (2022)
Keyphrases
- time of flight
- low memory
- depth map
- depth information
- multi view
- d scene
- high quality
- three dimensional
- depth images
- tof camera
- depth data
- stereo matching
- input image
- depth estimation
- depth video
- image sequences
- depth cameras
- view synthesis
- high resolution
- stereo pair
- real scenes
- super resolution
- infrared
- input data
- high accuracy
- image data
- face images
- scene geometry
- range data
- depth values
- single image
- depth estimates
- multiresolution
- pose estimation