Surface Normal Prediction on Hypersim benchmark leaderboard · SOTA2 Research

Surface Normal Prediction on Hypersim

-0.48Surface Normal Error (px)Blind guess

Evaluation Results

Method	Links
Blind guess2025.07		-0.48	-0.61	0.11
Gemini 2.0 Flash2025.07		-0.39	-0.04	0.02
Claude 3.5 Sonnet2025.07		-0.19	0.61	0.4
Gemini 1.5 Pro2025.07		-0.17	-0.57	0.04
GPT-4o2025.07		-0.14	0.57	0.4
Qwen2-VL2025.07		0.09	-0.07	0.02
o4-mini2025.07		0.22	0.61	0.46
Llama 3.22025.07		0.41	-0.42	0.22
DSINE2025.07		0.63	0.69	0.55
Omnidata2025.07		0.64	0.7	0.58
Lotus2025.07		0.64	0.7	0.54
Moge2025.07		0.64	0.69	0.56
Oracle2025.07		0.64	0.7	0.6
4M-212025.07		0.65	0.7	0.56
4M-212025.07		0.71	0.74	0.65
DSINE2025.07		0.76	0.79	0.73
Lotus2025.07		0.77	0.8	0.75
Omnidata2025.07		0.8	0.83	0.78
Moge2025.07		0.8	0.8	0.77