[ICCV 2023] Official implementation for "Label-Guided Knowledge Distillation for Continual Semantic Segmentation on 2D Images and 3D Point Clouds".
Official implementation of "Decoupling Continual Semantic Segmentation". Novel framework separating class-aware detection from class-agnostic segmentation for effective continual learning.
[Neural Networks 2023] The official repository of Neural Networks Journal "Subspace Distillation for Continual Learning"
A novel framework that uses foreground class prototypes along with dark experience replay to alleviate both catastrophic forgetting and background shift in semantic segmentation (where old classes are having ground truth background in future tasks).
To associate your repository with the continual-semantic-segmentation topic, visit your repo's landing page and select "manage topics."