Text-to-Image Generation on DrawBench (Fidelity Metrics) benchmark leaderboard · SOTA2 Research

Text-to-Image Generation on DrawBench (Fidelity Metrics)

93.13Spatial Fidelity (Human)MIGC

Evaluation Results

Method	Links
MIGC2024.02		93.13	69.38	79	97.5	67.76	67.5
GLIGEN2024.02		78.8	61.25	51	48.2	44.08	55.9
Multi-D2024.02		55.63	10.63	18.5	65.5	17.76	36.18
TFLCG2024.02		53.13	9.38	35	60	15.79	31.58
Box-D2024.02		50	11.88	28.5	57.5	9.21	39.47
AAE2024.02		23.13	—	—	51.5	—	30.92
SD1.42024.02		13.3	—	—	57.52	—	23.7
Struc-D2024.02		13.12	—	—	56.5	—	30.26