Text-to-3D Generation on Objaverse 1.0 (test) benchmark leaderboard · SOTA2 Research

Text-to-3D Generation on Objaverse 1.0 (test)

81.6CLIP ScoreGround Truth Images

Evaluation Results

Method	Links
Ground Truth Images2023.06		81.6	—	32.7	55.1	64.3
Shap-E (STF)2023.06		80.4	37.2	20.3	39.7	48.7
Shap-E (NeRF)2023.06		79.4	48.7	19	37.7	46.8
Shap-E (STF)2023.06		79.1	35.5	20	38.8	47.3
Shap-E (NeRF)2023.06		78.1	48.2	18.3	35.1	43.5
Point-E (Text-to-3D)2023.06		75.6	32.8	12.4	28.1	36.9
S. Diff. (CNet) + Point-E (Im-to-3D)2023.06		74.6	53.3	12.4	26.2	33.8
S. Diff. (LoRA) + Point-E (Im-to-3D)2023.06		74.4	53.7	11.6	24.6	31.4
S. Diff. (CNet) + Point-E (Im-to-3D)2023.06		73.6	54.7	11	23.4	30
S. Diff. (LoRA) + Point-E (Im-to-3D)2023.06		73.6	54.7	11	23.4	30
Point-E (Text-to-3D)2023.06		72.4	36.1	6	16.2	22.4