PathDiff jointly trains a latent diffusion model on unpaired text-image and mask-image histopathology datasets, enabling text-, mask-, or both-conditioned generation with reported gains in fidelity and downstream tasks.
Diffusion models beat gans on image synthesis
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
-
PathDiff: Histopathology Image Synthesis with Unpaired Text and Mask Conditions
PathDiff jointly trains a latent diffusion model on unpaired text-image and mask-image histopathology datasets, enabling text-, mask-, or both-conditioned generation with reported gains in fidelity and downstream tasks.