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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Recent advances in generative deep learning have enabled the creation ofhigh-quality synthetic images in text-to-image generation. Prior work showsthat fine-tuning a pretrained diffusion model on ImageNet and generatingsynthetic training images from the finetuned model can enhance an ImageNetclassifier's performance. However, performance degrades as synthetic imagesoutnumber real ones. In this paper, we explore whether generative fine-tuningis essential for this improvement and whether it is possible to further scaleup training using more synthetic data. We present a new framework leveragingoff-the-shelf generative models to generate synthetic training images,addressing multiple challenges: class name ambiguity, lack of diversity innaive prompts, and domain shifts. Specifically, we leverage large languagemodels (LLMs) and CLIP to resolve class name ambiguity. To diversify images, wepropose contextualized diversification (CD) and stylized diversification (SD)methods, also prompted by LLMs. Finally, to mitigate domain shifts, we leveragedomain adaptation techniques with auxiliary batch normalization for syntheticimages. Our framework consistently enhances recognition model performance withmore synthetic data, up to 6x of original ImageNet size showcasing thepotential of synthetic data for improved recognition models and strongout-of-domain generalization.

Diversify, Don't Fine-Tune: Scaling Up Visual Recognition Training with Synthetic Images