Multistage Self-Supervised Augmented Consecutive Knowledge Distillation (MSACKD)
Multistage self-supervised knowledge distillation for efficient multimodal learning.
CVPR Workshops, 2024.
Note
Distils a large multimodal model into a smaller one in successive stages, so that each stage recovers the representations the previous stage lost.
Reference
Multistage Self-Supervised Augmented Consecutive Knowledge Distillation (MSACKD) CVPR Workshops, 2024. https://scholar.google.com/scholar?q=%22Multistage%20Self-Supervised%20Augmented%20Consecutive%20Knowledge%20Distillation%20(MSACKD)%22
The publisher record holds the author list.
Details
- TypeConference paper
- AreasMedical imaging and computer vision, Foundation models for biology and medicine
More from the index
- PreviousSingle Stage Adaptive Multi-Attention Network for Image RestorationIEEE Transactions on Image Processing
- NextExploring Multiple Instance Learning (MIL): A brief surveyExpert Systems with Applications