🏢 CompVis @ LMU Munich, MCML
CleanDIFT: Diffusion Features without Noise
·3337 words·16 mins·
loading
·
loading
AI Generated
🤗 Daily Papers
Computer Vision
Image Generation
🏢 CompVis @ LMU Munich, MCML
CleanDIFT revolutionizes diffusion feature extraction by leveraging clean images and a lightweight fine-tuning method, significantly boosting performance across various tasks without noise or timestep…