Evaluasi Kualitas Generasi Objek 3D Menggunakan Hyperscore pada Subset Benchmark

Nurul Humaera B; Farida Farida; Amhar Akram Haidar

doi:10.55338/jumin.v7i5.9381

Authors

Nurul Humaera B Institut Teknologi dan Bisnis Bina Adinata
Farida Farida Institut Teknologi dan Bisnis Bina Adinata
Amhar Akram Haidar Institut Teknologi dan Bisnis Bina Adinata

DOI:

https://doi.org/10.55338/jumin.v7i5.9381

Keywords:

Text-to-3D, Generative AI, Hyperscore, 3D Object Generation, Subset evaluation

Abstract

Akselerasi teknologi AI generatif teks-ke-3D (Text-to-3D) menjadi salah satu pendekatan untuk meningkatkan efisiensi dalam pembuatan aset 3D. Evaluasi menggunakan full Benchmark membutuhkan sumber daya yang relatif tinggi akibat biaya, kuota API, dan kebutuhan komputasi, sehingga diperlukan pendekatan subset Benchmark yang lebih efisien. Penelitian ini bertujuan untuk mengevaluasi kualitas hasil generasi objek 3D menggunakan skema subset evaluation pada Benchmark Hyperscore. Penelitian ini adalah analisis empiris awal dengan ruang lingkup terbatas pada 24 prompt dan dihasilkan menggunakan platform Meshy.AI. Generasi objek 3D dilakukan menggunakan platform Meshy.AI, kemudian dievaluasi menggunakan empat dimensi Hyperscore, yaitu Alignment, Geometry, Texture, dan Overall Quality. Hasil evaluasi menunjukkan adanya variasi kualitas antarkategori prompt. Kategori Basic memperoleh skor Overall Quality tertinggi sebesar 7,171, sedangkan kategori Action memperoleh skor terendah sebesar 5,513. Kategori Spatial juga menunjukkan variasi hasil yang relatif besar pada dimensi Geometry. Temuan ini menunjukkan bahwa skema subset evaluation berpotensi menjadi alternatif evaluasi yang lebih efisien untuk penelitian Text-to-3D tanpa menghilangkan kemampuan untuk menggambarkan variasi kualitas antarkategori. Namun, penelitian ini belum membandingkan hasil subset secara langsung dengan keseluruhan 160 prompt, sehingga kesetaraan subset terhadap Benchmark penuh belum dapat disimpulkan. Hasil penelitian ini dapat menjadi dasar untuk kajian lebih lanjut mengenai strategi pemilihan subset benchmark dalam evaluasi generasi Text-to-3D.

Downloads

Download data is not yet available.

References

R. Rombach, A. Blattmann, and D. Lorenz, “High-Resolution Image Synthesis with Latent Diffusion Models,” IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pp. 10674–10685, 2022, doi: 10.1109/cvpr52688.2022.01042.

B. Poole, A. Jain, J. T. Barron, and B. Mildenhall, “DREAMFUSION: TEXT-TO-3D USING 2D DIFFUSION,” in International Conference on Learning Representations (ICLR), 2023, p. 2209.14988, [Online]. Available: https://openreview.net/forum?id=FjNys5c7VyY.

C. Lin et al., “Magic3D: High-Resolution Text-to-3D Content Creation,” IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2023, doi: 10.48550/arXiv.2211.10440.

T. Hunyuan, “Hunyuan3D 1.0: A Unified Framework for Text-to-3D and Image-to-3D Generation,” arXiv, 2025, doi: 10.48550/arXiv.2411.02293.

A. T. Prakash, S. Shaji, and S. Raghunath, “A SURVEY ON TEXT TO 3D MODEL GENERATOR,” IJRTI2501071 Int. J. Res. Trends Innov., vol. 10, no. 1, pp. 585–589, 2025, [Online]. Available: https://www.ijrti.org/papers/IJRTI2501071.pdf.

R. Ziatdinov and R. Nabiyev, “Generating Digital Models Using Text-to-3D and Image-to-3D Prompts: Critical Case Study,” Proc. - 2025 Int. Russ. Smart Ind. Conf. SmartIndustryCon 2025, pp. 719–724, 2025, doi: 10.1109/SmartIndustryCon65166.2025.10986113.

B. Mildenhall, J. T. Barron, R. Ramamoorthi, and R. Ng, “NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis,” Commun. ACM, vol. 65, no. 1, pp. 99–106, 2021, doi: 10.1145/3503250.

R. Liu, R. Wu, B. Van Hoorick, P. Tokmakov, S. Zakharov, and C. Vondrick, “Zero-1-to-3: Zero-shot One Image to 3D Object,” Proc. IEEE Int. Conf. Comput. Vis., pp. 9264–9275, 2023, doi: 10.1109/ICCV51070.2023.00853.

A. Nichol, H. Jun, P. Dhariwal, P. Mishkin, and M. Chen, “Point-E: A System for Generating 3D Point Clouds from Complex Prompts,” arXiv, 2022, doi: 10.48550/arXiv.2212.08751.

H. Jun and A. Nichol, “Shap-E: Generating Conditional 3D Implicit Functions,” IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2023, doi: 10.48550/arxiv.2305.02463.

R. Chen, Y. Chen, N. Jiao, and K. Jia, “Fantasia3D: Disentangling Geometry and Appearance for High-quality Text-to-3D Content Creation,” Proc. IEEE Int. Conf. Comput. Vis., pp. 22189–22199, 2023, doi: 10.1109/ICCV51070.2023.02033.

J. Tang, J. Ren, H. Zhou, Z. Liu, and Z. Gang, “DREAMGAUSSIAN: GENERATIVE GAUSSIAN SPLAT-TING FOR EFFICIENT 3D CONTENT CREATION,” in The International Conference on Learning Representations (ICLR), 2024, pp. 1–18, doi: 10.48550/arXiv.2309.16653.

B. Pecher, M. Spiegel, R. Belanec, and J. Cegin, “Revisiting Prompt Sensitivity in Large Language Models for Text,” ArXiv, abs, p. 2602.04297, 2026, doi: 10.48550/arXiv.2602.04297.

A. Radford et al., “Learning Transferable Visual Models From Natural Language Supervision,” Int. Conf. Mach. Learn. PMLR, 2021, doi: 10.48550/arxiv.2103.00020.

Y. Zhang, B. Cui, Q. Yang, Z. Li, and Y. Xu, “Benchmarking and Learning Multi-Dimensional Quality Evaluator for Text-To-3D Generation,” Int. Conf. Comput. Vis., pp. 18563–18574, 2025, doi: 10.1109/ICCV51701.2025.01725.

S. Maiti, L. Agapito, and F. Kokkinos, “Gen3DEval: Using vLLMs for Automatic Evaluation of Generated 3D Objects,” IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pp. 18552–18562, 2025, doi: 10.1109/cvpr52734.2025.01729.

Y. Zhang et al., “3DGen-Bench: Comprehensive Benchmark Suite for 3D Generative Models,” IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2025, doi: 10.48550/arXiv.2503.21745.

X. Fei, A. Swaminathan, and P. Favaro, “A Quantitative Evaluation of Score Distillation Sampling Based Text-to-3D,” arXiv, 2024, doi: 10.48550/arxiv.2402.18780.

Y. He et al., “T3Bench: Benchmarking Current Progress in Text-to-3D Generation,” 2024, doi: 10.48550/arXiv.2310.02977.

T. Wu et al., “GPT-4V(ision) is a Human-Aligned Evaluator for Text-to-3D Generation,” Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2024, doi: 10.48550/arXiv.2401.04092.

M. Sawant, “Advancements in AI for Text-to-3D Model Generation : A Comparative Study of Meshy.ai and LumaLabs.ai Genie,” Int. J. Adv. Comput. Eng. Commun. Technol., vol. 14, no. 1, pp. 485–489, 2025, doi: 10.65521/ijacect.v14i1.563.

Y. Zhou et al., “3DGCQA: A Quality Assessment Database for 3D AI-Generated Contents,” ICASSP, IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., no. c, 2025, doi: 10.1109/ICASSP49660.2025.10889921.

S. Duggal et al., “Eval3D: Interpretable and Fine-Grained Evaluation for 3D Generation,” IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pp. 13326–13336, 2025, doi: 10.48550/arXiv.2504.18509.

B. Kerbl et al., “3D Gaussian Splatting for Real-Time Radiance Field Rendering,” ACM Trans. Graph., vol. 42, no. 4, pp. 1–14, 2023, doi: 10.1145/3592433.

O. Michel, R. Bar-On, R. Liu, S. Benaim, and R. Hanocka, “Text2Mesh : Text-Driven Neural Stylization for Meshes,” Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., pp. 13492–13502, 2022, doi: 10.1109/CVPR52688.2022.01313.