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/

Global: Zendy Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Classifiers and generators have long been separated. We break down thisseparation and showcase that conventional neural network classifiers cangenerate high-quality images of a large number of categories, being comparableto the state-of-the-art generative models (e.g., DDPMs and GANs). We achievethis by computing the partial derivative of the classification loss functionwith respect to the input to optimize the input to produce an image. Since itis widely known that directly optimizing the inputs is similar to targetedadversarial attacks incapable of generating human-meaningful images, we proposea mask-based stochastic reconstruction module to make the gradientssemantic-aware to synthesize plausible images. We further propose aprogressive-resolution technique to guarantee fidelity, which producesphotorealistic images. Furthermore, we introduce a distance metric loss and anon-trivial distribution loss to ensure classification neural networks cansynthesize diverse and high-fidelity images. Using traditional neural networkclassifiers, we can generate good-quality images of 256$\times$256 resolutionon ImageNet. Intriguingly, our method is also applicable to text-to-imagegeneration by regarding image-text foundation models as generalizedclassifiers.  Proving that classifiers have learned the data distribution and are ready forimage generation has far-reaching implications, for classifiers are much easierto train than generative models like DDPMs and GANs. We don't even need totrain classification models because tons of public ones are available fordownload. Also, this holds great potential for the interpretability androbustness of classifiers. Project page is at\url{https://classifier-as-generator.github.io/}.

Traditional Classification Neural Networks are Good Generators: They are Competitive with DDPMs and GANs