notesum.ai

Published at November 22

High-Resolution Image Synthesis via Next-Token Prediction

cs.CV

cs.AI

Released Date: November 22, 2024

Authors: Dengsheng Chen¹, Jie Hu¹, Tiezhu Yue¹, Xiaoming Wei¹

Aff.: ¹Meituan, Beijing, China

Arxiv: http://arxiv.org/abs/2411.14808v1

			GenEval							T2I-CompBench++
Model	NTP	#Params	Overall	Single Obj.	Two Obj.	Counting	Colors	Position	Color Attr.	Color	Shape	Texture
Small model size
PixArt- $\alpha$ [21]	$\times$	0.6B	0.48	0.98	0.50	0.44	0.80	0.08	0.07	0.4232	0.3764	0.4808
SD v1.x [93]	$\times$	0.9B	0.43	0.97	0.38	0.35	0.76	0.04	0.06	0.3765	0.3576	0.4156
SD v2.x [93]	$\times$	0.9B	0.50	0.98	0.51	0.44	0.85	0.07	0.17	0.5065	0.4221	0.4922
SD 3.0 [36]	$\times$	1.0B	0.58	0.97	0.72	0.52	0.78	0.16	0.34	-	-	-
Show-o [124]	$\checkmark$	1.3B	0.53	0.95	0.52	0.49	0.82	0.11	0.28	-	-	-
LDM [93]	$\times$	1.4B	0.37	0.92	0.29	0.23	0.70	0.02	0.05	-	-	-
Hunyuan-DiT [64]	$\times$	1.5B	0.57	0.96	0.67	0.59	0.83	0.11	0.26	0.6565	0.3577	0.4718
Mainstream model size
Lumina-T2I [136]	$\times$	2.0B	0.39	0.88	0.34	0.31	0.67	0.05	0.09	0.4081	0.3008	0.4071
SD 3.0 [36]	$\times$	2.0B	0.62	0.98	0.74	0.63	0.67	0.34	0.36	0.8132	0.5885	0.7334
D-JEPA $\cdot$ T2I	$\checkmark$	2.6B	0.66	0.99	0.80	0.59	0.87	0.22	0.47	0.7585	0.5036	0.6355
SDXL [83]	$\times$	2.6B	0.55	0.98	0.74	0.39	0.85	0.15	0.23	0.5879	0.4687	0.5299
LlamaGen [109]	$\checkmark$	3.1B	0.32	0.71	0.34	0.21	0.58	0.07	0.04	-	-	-
SD 3.0 [36]	$\times$	4.0B	0.64	0.96	0.80	0.65	0.73	0.33	0.37	-	-	-
DALL $\cdot$ E 2 [90]	$\times$	4.2B	0.52	0.94	0.66	0.49	0.77	0.10	0.19	-	-	-
Extensive model size
Chameleon [111]	$\checkmark$	7.0B	0.39	-	-	-	-	-	-	-	-	-
Transfusion [135]	$\checkmark$	7.3B	0.63	-	-	-	-	-	-	-	-	-
SD 3.0 [36]	$\times$	8.0B	0.68	0.98	0.84	0.66	0.74	0.40	0.43	-	-	-
Emu3 [122]	$\checkmark$	8.0B	0.54	0.98	0.71	0.34	0.81	0.17	0.21	-	-	-
Fluid [37]	$\checkmark$	10.5B	0.69	0.96	0.83	0.63	0.80	0.39	0.51	-	-	-
FLUX.1.dev [59]	$\times$	12.0B	-	-	-	-	-	-	-	0.7407	0.5718	0.6922
DALL $\cdot$ E 3 [12]	$\times$	-	0.67	0.96	0.87	0.47	0.83	0.43	0.45	0.7785	0.6205	0.7036
Midjourney v6 [75]	$\times$	-	0.63	0.96	0.81	0.56	0.83	0.22	0.42	0.7503	0.6885	0.6101