Fine-tuning a CLIP image encoder with a diffusion reconstruction loss through a frozen unCLIP generator improves its fine-grained visual understanding on several benchmarks, while degrading coarse-grained classification.
Diffusion feedback helps CLIP see better
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
un$^2$CLIP: Improving CLIP's Visual Detail Capturing Ability via Inverting unCLIP
Fine-tuning a CLIP image encoder with a diffusion reconstruction loss through a frozen unCLIP generator improves its fine-grained visual understanding on several benchmarks, while degrading coarse-grained classification.