Image Generation On Textatlaseval

StyledTextSynth Clip Score

StyledTextSynth FID

StyledTextSynth OCR (Accuracy)

StyledTextSynth OCR (Cer)

StyledTextSynth OCR (F1 Score)

TextScenesHQ Clip Score

TextScenesHQ FID

TextScenesHQ OCR (Accuracy)

TextScenesHQ OCR (Cer)

TextScenesHQ OCR (F1 Score)

TextVisionBlend Clip Score

TextVisionBlend FID

TextVisionBlend OCR (Accuracy)

TextVisionBlend OCR (Cer)

TextVsionBlend OCR (F1 Score)

Results

Performance results of various models on this benchmark

																Paper Title	Repository
Dalle3	0.2938	90.70	30.58	0.78	38.25	0.3367	86.73	69.26	-	51.63	0.1938	153.21	8.38	0.93	7.94	-	-
Grok3	0.2938	80.33	15.82	0.73	21.40	0.3197	-	35.07	0.57	37.94	0.1697	-	41.54	0.57	44.22	-	-
SD3.5 Large	0.2849	71.09	27.21	0.73	33.86	0.2363	64.44	19.03	0.73	24.45	0.1846	118.85	14.55	0.88	16.25	-	-
PixArt-Sigma	0.2764	82.83	0.42	0.90	0.62	0.2347	72.62	0.34	0.91	0.53	0.1891	81.29	2.40	0.83	1.57	PixArt-Σ: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation
Infinity-2B	0.2727	84.95	0.80	0.93	1.42	0.2346	71.59	1.06	0.88	1.74	0.1979	95.69	2.98	0.83	3.44	Infinity-MM: Scaling Multimodal Performance with Large-Scale and High-Quality Instruction Data
TextDiffuser2	0.2510	114.31	0.76	0.99	1.46	0.2252	84.10	0.66	0.96	1.25	-	-	-	-	-	TextDiffuser-2: Unleashing the Power of Language Models for Text Rendering	-
Anytext	0.2501	117.71	0.35	0.98	0.66	0.2174	101.32	0.42	0.95	0.8	-	-	-	-	-	AnyText: Multilingual Visual Text Generation And Editing

0 of 7 row(s) selected.