CATS: Classification Aided Transport Splatting for Energy-Efficient 3D Reconstruction

• Paulo Abner A Mesquita Instituto Atlântico
• Leonardo Rocha Instituto Atlântico

DOI: https://doi.org/10.5753/semish.2026.20869

Resumo

A demanda por reconstrução 3D impõe altos custos energéticos, exigindo estratégias de Green AI. Este trabalho propõe o método CATS, que utiliza Transporte Ótimo (OT) sobre embeddings semânticos do modelo CLIP para selecionar as imagens mais informativas para o 3D Gaussian Splatting (3DGS). Monitorando o consumo (Wh) e a emissão de CO2 via biblioteca CodeCarbon, comparamos o OT contra a seleção aleatória em diferentes volumes de dados. O OT reduz o consumo em datasets densos eliminando redundâncias, com trade-off de aproximadamente 5 dB de PSNR. Em cenários de escassez (10–30%), a baixa conectividade geométrica torna o método menos robusto, gerando emissões até 40% superiores à seleção aleatória.

Referências

Arjovsky, M., Chintala, S., and Bottou, L. (2017). Wasserstein generative adversarial networks. In International conference on machine learning, pages 214–223. PMLR.

Courty, B., Schmidt, V., Goyal-Kamal, MarionCoutarel, Feld, B., Lecourt, J., LiamConnell, SabAmine, inimaz, supatomic, Léval, M., Blanche, L., Cruveiller, A., ouminasara, Zhao, F., Joshi, A., Bogroff, A., Saboni, A., de Lavoreille, H., Laskaris, N., Abati, E., Blank, D., Wang, Z., Catovic, A., alencon, Bauer, C., Lucas-Otavio, JPW, and MinervaBooks (2024). mlco2/codecarbon: v2.4.1.

Cuturi, M. (2013). Sinkhorn distances: Lightspeed computation of optimal transport. In Advances in Neural Information Processing Systems (NeurIPS), pages 2292–2300.

Fischer, R. (2025). Ground-truthing ai energy consumption: Validating codecarbon against external measurements.

Furukawa, Y., Curless, B., Seitz, S. M., and Szeliski, R. (2010). Towards internet-scale multi-view stereo. In 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pages 1434–1441.

Just, H. A., Kang, F., Wang, T., Zeng, Y., Ko, M., Jin, M., and Jia, R. (2023). LAVA: Data valuation without pre-specified learning algorithms. In The Eleventh International Conference on Learning Representations.

Kerbl, B., Kopanas, G., Leimkühler, T., and Drettakis, G. (2023). 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 42(4).

Kessler, S., Le, T., and Nguyen, V. (2024). Sava: Scalable learning-agnostic data valuation.

Mildenhall, B., Srinivasan, P. P., Tancik, M., Barron, J. T., Ramamoorthi, R., and Ng, R. (2020). Nerf: Representing scenes as neural radiance fields for view synthesis.

Peyré, G., Cuturi, M., et al. (2019). Computational optimal transport: With applications to data science. Foundations and Trends® in Machine Learning, 11(5-6):355–607.

Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al. (2021). Learning transferable visual models from natural language supervision. In International Conference on Machine Learning (ICML), pages 8748–8763. PMLR.

Rubner, Y., Tomasi, C., and Guibas, L. J. (2000). The earth mover’s distance as a metric for image retrieval. International journal of computer vision, 40(2):99–121.

Schönberger, J. L. and Frahm, J.-M. (2016). Structure-from-motion revisited. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 4104–4113.

Strubell, E., Ganesh, A., and McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. In Korhonen, A., Traum, D., and Màrquez, L., editors, Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3645–3650, Florence, Italy. Association for Computational Linguistics.

Thengane, V., Zhu, X., Bouzerdoum, S., Phung, S. L., and Li, Y. (2025). Foundational models for 3d point clouds: A survey and outlook.