X-Oscar: A Progressive Framework for High-quality Text-guided 3D Animatable Avatar Generation

Yiwei Ma, Zhekai Lin, Jiayi Ji, Yijun Fan, Xiaoshuai Sun, Rongrong Ji
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:33826-33838, 2024.

Abstract

Recent advancements in automatic 3D avatar generation guided by text have made significant progress. However, existing methods have limitations such as oversaturation and low-quality output. To address these challenges, we propose X-Oscar, a progressive framework for generating high-quality animatable avatars from text prompts. It follows a sequential "Geometry→Texture→Animation" paradigm, simplifying optimization through step-by-step generation. To tackle oversaturation, we introduce Adaptive Variational Parameter (AVP), representing avatars as an adaptive distribution during training. Additionally, we present Avatar-aware Score Distillation Sampling (ASDS), a novel technique that incorporates avatar-aware noise into rendered images for improved generation quality during optimization. Extensive evaluations confirm the superiority of X-Oscar over existing text-to-3D and text-to-avatar approaches. Our anonymous project page: https://anonymous1440.github.io/.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-ma24g, title = {X-Oscar: A Progressive Framework for High-quality Text-guided 3{D} Animatable Avatar Generation}, author = {Ma, Yiwei and Lin, Zhekai and Ji, Jiayi and Fan, Yijun and Sun, Xiaoshuai and Ji, Rongrong}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {33826--33838}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/ma24g/ma24g.pdf}, url = {https://proceedings.mlr.press/v235/ma24g.html}, abstract = {Recent advancements in automatic 3D avatar generation guided by text have made significant progress. However, existing methods have limitations such as oversaturation and low-quality output. To address these challenges, we propose X-Oscar, a progressive framework for generating high-quality animatable avatars from text prompts. It follows a sequential "Geometry→Texture→Animation" paradigm, simplifying optimization through step-by-step generation. To tackle oversaturation, we introduce Adaptive Variational Parameter (AVP), representing avatars as an adaptive distribution during training. Additionally, we present Avatar-aware Score Distillation Sampling (ASDS), a novel technique that incorporates avatar-aware noise into rendered images for improved generation quality during optimization. Extensive evaluations confirm the superiority of X-Oscar over existing text-to-3D and text-to-avatar approaches. Our anonymous project page: https://anonymous1440.github.io/.} }
Endnote
%0 Conference Paper %T X-Oscar: A Progressive Framework for High-quality Text-guided 3D Animatable Avatar Generation %A Yiwei Ma %A Zhekai Lin %A Jiayi Ji %A Yijun Fan %A Xiaoshuai Sun %A Rongrong Ji %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-ma24g %I PMLR %P 33826--33838 %U https://proceedings.mlr.press/v235/ma24g.html %V 235 %X Recent advancements in automatic 3D avatar generation guided by text have made significant progress. However, existing methods have limitations such as oversaturation and low-quality output. To address these challenges, we propose X-Oscar, a progressive framework for generating high-quality animatable avatars from text prompts. It follows a sequential "Geometry→Texture→Animation" paradigm, simplifying optimization through step-by-step generation. To tackle oversaturation, we introduce Adaptive Variational Parameter (AVP), representing avatars as an adaptive distribution during training. Additionally, we present Avatar-aware Score Distillation Sampling (ASDS), a novel technique that incorporates avatar-aware noise into rendered images for improved generation quality during optimization. Extensive evaluations confirm the superiority of X-Oscar over existing text-to-3D and text-to-avatar approaches. Our anonymous project page: https://anonymous1440.github.io/.
APA
Ma, Y., Lin, Z., Ji, J., Fan, Y., Sun, X. & Ji, R.. (2024). X-Oscar: A Progressive Framework for High-quality Text-guided 3D Animatable Avatar Generation. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:33826-33838 Available from https://proceedings.mlr.press/v235/ma24g.html.

Related Material