SIR-Net: Self-Supervised Transfer for Inverse Rendering via Deep Feature Fusion and Transformation From a Single Image.
Tianteng BiJunjie MaYue LiuDongdong WengYongtian WangPublished in: IEEE Access (2020)
Keyphrases
- single image
- feature fusion
- d scene
- feature extraction
- multiple features
- shape recovery
- multiple images
- light source
- data reduction
- high quality
- super resolution
- subspace learning
- multi sensor
- intrinsic images
- fusion algorithm
- computer vision
- transfer learning
- natural images
- high dimensional
- d objects
- preprocessing
- fusion method
- data analysis
- scene geometry
- recognition rate