English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Denoising Diffusion Probabilistic Model (DDPM) is able to make flexibleconditional image generation from prior noise to real data, by introducing anindependent noise-aware classifier to provide conditional gradient guidance ateach time step of denoising process. However, due to the ability of classifierto easily discriminate an incompletely generated image only with high-levelstructure, the gradient, which is a kind of class information guidance, tendsto vanish early, leading to the collapse from conditional generation processinto the unconditional process. To address this problem, we propose two simplebut effective approaches from two perspectives. For sampling procedure, weintroduce the entropy of predicted distribution as the measure of guidancevanishing level and propose an entropy-aware scaling method to adaptivelyrecover the conditional semantic guidance. For training stage, we propose theentropy-aware optimization objectives to alleviate the overconfident predictionfor noisy data.On ImageNet1000 256x256, with our proposed sampling scheme andtrained classifier, the pretrained conditional and unconditional DDPM model canachieve 10.89% (4.59 to 4.09) and 43.5% (12 to 6.78) FID improvementrespectively. The code is available at https://github.com/ZGCTroy/ED-DPM.

Entropy-driven Sampling and Training Scheme for Conditional Diffusion Generation