PODIA-3D: Domain Adaptation of 3D Generative Model Across Large Domain Gap Using Pose-Preserved Text-to-Image Diffusion.
Gwanghyun KimJi Ha JangSe Young ChunPublished in: ICCV (2023)
Keyphrases
- generative model
- domain adaptation
- bayesian framework
- semi supervised
- cross domain
- image classification
- image features
- probabilistic model
- multiscale
- input image
- prior knowledge
- target domain
- image retrieval
- image segmentation
- labeled data
- multiple sources
- transfer learning
- domain specific
- expectation maximization
- keywords
- image representation
- topic models
- information retrieval
- sentiment classification
- semi supervised learning
- high resolution
- object recognition
- unlabeled data
- feature points
- data sets