Single-image driven 3d viewpoint training data augmentation for effective wine label recognition.
Yueh-Cheng HuangHsinYi ChenCheng-Jui HungJen-Hui ChuangJenq-Neng HwangPublished in: CoRR (2024)
Keyphrases
- single image
- viewpoint
- training data
- multiple images
- shape recovery
- d scene
- light source
- object recognition
- image pairs
- shape from shading
- partially occluded
- vanishing points
- training set
- super resolution
- spherical harmonics
- feature extraction
- intrinsic image decomposition
- d objects
- class labels
- multiscale
- high quality
- indoor scenes
- information from multiple images