Segment scene with new objects without forgetting old ones using visual prompt tuning with ECLIPSE
Segment scene with new objects without forgetting old ones using visual prompt tuning with ECLIPSE
ECLIPSE: Efficient Continual Learning in Panoptic Segmentation with Visual Prompt Tuning
arXiv paper abstract https://arxiv.org/abs/2403.20126
arXiv PDF paper https://arxiv.org/pdf/2403.20126.pdf
Panoptic segmentation, combining semantic and instance segmentation ... real-world ... necessitates continual learning, where models adapt to new classes (plasticity) ... without forgetting old ones (catastrophic forgetting).
... continual segmentation ... often rely on distillation strategies like knowledge distillation and pseudo-labeling ... result in increased training complexity and computational overhead.
... introduce a novel and efficient method for continual panoptic segmentation based on Visual Prompt Tuning, dubbed ECLIPSE.
... involves freezing the base model parameters and fine-tuning only a small set of prompt embeddings, addressing both catastrophic forgetting and plasticity and significantly reducing the trainable parameters.
To mitigate ... error propagation and semantic drift in continual segmentation, ... propose logit manipulation to effectively leverage common knowledge across the classes.
... demonstrate the superiority of ECLIPSE, notably its robustness against catastrophic forgetting and its reasonable plasticity, achieving a new state-of-the-art ...
Please like and share this post if you enjoyed it using the buttons at the bottom!
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
Comments