MVSFormer: Multi-View Stereo with Pre-trained Vision Transformers and Temperature-based Depth.
Chenjie CaoXinlin RenYanwei FuPublished in: CoRR (2022)
Keyphrases
- multi view stereo
- pre trained
- depth map
- depth estimation
- multi view
- depth information
- stereo algorithms
- training data
- stereo matching
- d scene
- stereo vision
- high quality
- computer vision
- vision system
- three dimensional
- structure from motion
- training examples
- stereo reconstruction
- view synthesis
- image sequences
- super resolution
- control signals
- input image
- real scenes
- high resolution
- visual hull
- surface reconstruction
- point cloud
- real time
- multiple views
- low resolution
- dynamic scenes
- stereo pair
- high fidelity
- depth discontinuities
- image space
- d objects