11 papers
Distill Once, Adapt Life-Long: Exploring Dataset Distillation for Continual Test-Time Adaptation
Hyun-Kurl Jang, Jihun Kim, Hyeokjun Kweon +1
Continual Test-Time Adaptation (CTTA) aims to maintain model performance under evolving target domains by adapting online without labeled data. However, practical deployments often…
FrozenDrive: Zero-Shot Text-Guided Driving Scene Generation and Data Augmentation with Parameter-Free Frozen Diffusion Model
Yuhwan Jeong, Hyeonseong Kim, Daehyun We +5
Synthetic data for autonomous driving is surging, powered by diffusion models that promise scalable scene generation. Yet key obstacles remain, as enforcing multi-view and temporal…
Bootstrapping Video Semantic Segmentation Model via Distillation-assisted Test-Time Adaptation
Jihun Kim, Hoyong Kwon, Hyeokjun Kweon +1
Fully supervised Video Semantic Segmentation (VSS) relies heavily on densely annotated video data, limiting practical applicability. Alternatively, applying pre-trained Image Seman…
DC-TTA: Divide-and-Conquer Framework for Test-Time Adaptation of Interactive Segmentation
Jihun Kim, Hoyong Kwon, Hyeokjun Kweon +2
Interactive segmentation (IS) allows users to iteratively refine object boundaries with minimal cues, such as positive and negative clicks. While the Segment Anything Model (SAM) h…
Label-Free Cross-Task LoRA Merging with Null-Space Compression
Wonyoung Lee, Wooseong Jeong, Kuk-Jin Yoon
Model merging combines independently fine-tuned checkpoints without joint multi-task training. In the era of foundation-model, fine-tuning with Low-Rank Adaptation (LoRA) is preval…
Preference-Aligned LoRA Merging: Preserving Subspace Coverage and Addressing Directional Anisotropy
Wooseong Jeong, Wonyoung Lee, Kuk-Jin Yoon
Merging multiple Low-Rank Adaptation (LoRA) modules is promising for constructing general-purpose systems, yet challenging because LoRA update directions span different subspaces a…