Signal Learning Lab

Test-Time Adaptation (TTA) adapts a pre-trained model to new data distributions during inference, without accessing the original training data.

• (New) "D-TPT (Test-Time Prompt Tuning) for Vision-Language Models"
  https://arxiv.org/abs/2510.09473

• (New) "When Test-Time Adaptation Meets Self-Supervised Models"
  https://arxiv.org/abs/2506.23529

• [ICML'25]
  "Ranked EM-based Continual TTA"
  https://arxiv.org/abs/2505.16441

Semantic Segmentation is the task of clustering parts of an image together which belong to the same object class.

• [NeurIPS'23]
  https://arxiv.org/abs/2310.18640

• [CVPR'22] https://arxiv.org/abs/2111.14173

Continual Learning is a model learning a large number of tasks sequentially without forgetting knowledge obtained from the preceding tasks.

• [WACV'25]
  https://arxiv.org/abs/2403.11537

• [ACCV'24] https://arxiv.org/abs/2305.05175

Domain Adaption aims to adapt models from a labeled source domain to a different but target domain without labels.

• https://arxiv.org/abs/2403.11582

• [CVPR'24]
  https://arxiv.org/abs/2403.09359

• [ECCV'22] https://arxiv.org/abs/2111.13353

• [CVPR'21] https://arxiv.org/abs/2011.09230 

Knowledge Distillation extracts pivtoal knowledge from a teacher network to guide the learning of a student network.

• [ICCV'23]
  https://arxiv.org/abs/2206.01186

• [CVPR'23] https://arxiv.org/abs/2205.15531

• [ICCV'21] https://arxiv.org/abs/2009.08825

Self-supervised Learning for leveraging training data without supervision signals for Classification and Detection.

• https://arxiv.org/abs/2003.02692