AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Image Captions are Natural Prompts for Training Data Synthesis

Shiye Lei, Hao Chen, Sen Zhang, Bo Zhao, Dacheng Tao

IJCV 2025 · Image generation › Text-to-image diffusion · approach: Personalized and region-controlled generation · 24 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace basic prompts with real-image captions (CiP)Stable Diffusion synthetic data · ImageNet-1K valtop-1 accuracy: 45.23 -> 54.06Improved
generate 10x more synthetic data with CiPCiP · ImageNet-1K valtop-1 accuracy: 54.06 -> 62.14Improved
use BLIP-2 captions instead of ViT-GPT2 (ablation)CiP · ImageNettetop-1 accuracy: 77.0 -> 79.0Improved
use real-image captions instead of zero-shot captions of synthetic images (ablation)CiP · ImageNette (gs=1.5)top-1 accuracy: 67.8 -> 79.0Improved
zero-shot CiP at large guidance scale (ablation)Basic prompts · ImageNette (gs=7.5)top-1 accuracy: 45.8 -> 49.2Improved
polish captions with Vicuna-13B (ablation)CiP (BLIP-2) · ImageNette (gs=1.5)top-1 accuracy: 79.0 -> 66.8Hurt
use CiP instead of LLM-summarized captionsLLM-Sum · ImageNet-1K valtop-1 accuracy: 53.27 -> 54.06Improved
LLM-generated captions instead of basic promptsBasic prompts · ImageNet-1K valtop-1 accuracy: 45.23 -> 43.13Hurt
use CiP instead of basic prompts with ViT-B/16ViT-B/16 from scratch · ImageNet-1K valtop-1 accuracy: 44.00 -> 53.04Improved
scale CiP synthetic data 10x for OoD robustnessCiP · ImageNet-A (200 classes)top-1 accuracy: 5.19 -> 10.44Improved
add CiP synthetic data to real dataResNet-50 on ImageNette · ImageNettetop-1 accuracy: +3.0%Improved
add CiP synthetic data to real dataResNet-50 on ImageNet-100 · ImageNet-100top-1 accuracy: +1.18%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy