ClustViT modifies ViT backbones with a clustering-based token merger guided by segmentation pseudo-labels and a detail regenerator to cut GFLOPs and speed up inference while keeping segmentation accuracy comparable on three datasets.
GTP-ViT: Efficient Vision Transformers via Graph-Based Token Propagation,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
ClustViT: Clustering-based Token Merging for Semantic Segmentation
ClustViT modifies ViT backbones with a clustering-based token merger guided by segmentation pseudo-labels and a detail regenerator to cut GFLOPs and speed up inference while keeping segmentation accuracy comparable on three datasets.