Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2504.11447 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912330604871680 |
|---|---|
| author | Zhao, An Zhang, Shengyuan Yang, Ling Li, Zejian Wu, Jiale Xu, Haoran Wei, AnYang GU, Perry Pengyun Sun, Lingyun |
| author_facet | Zhao, An Zhang, Shengyuan Yang, Ling Li, Zejian Wu, Jiale Xu, Haoran Wei, AnYang GU, Perry Pengyun Sun, Lingyun |
| contents | The application of diffusion models in 3D LiDAR scene completion is limited due to diffusion's slow sampling speed. Score distillation accelerates diffusion sampling but with performance degradation, while post-training with direct policy optimization (DPO) boosts performance using preference data. This paper proposes Distillation-DPO, a novel diffusion distillation framework for LiDAR scene completion with preference aligment. First, the student model generates paired completion scenes with different initial noises. Second, using LiDAR scene evaluation metrics as preference, we construct winning and losing sample pairs. Such construction is reasonable, since most LiDAR scene metrics are informative but non-differentiable to be optimized directly. Third, Distillation-DPO optimizes the student model by exploiting the difference in score functions between the teacher and student models on the paired completion scenes. Such procedure is repeated until convergence. Extensive experiments demonstrate that, compared to state-of-the-art LiDAR scene completion diffusion models, Distillation-DPO achieves higher-quality scene completion while accelerating the completion speed by more than 5-fold. Our method is the first to explore adopting preference learning in distillation to the best of our knowledge and provide insights into preference-aligned distillation. Our code is public available on https://github.com/happyw1nd/DistillationDPO. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_11447 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Diffusion Distillation With Direct Preference Optimization For Efficient 3D LiDAR Scene Completion Zhao, An Zhang, Shengyuan Yang, Ling Li, Zejian Wu, Jiale Xu, Haoran Wei, AnYang GU, Perry Pengyun Sun, Lingyun Computer Vision and Pattern Recognition The application of diffusion models in 3D LiDAR scene completion is limited due to diffusion's slow sampling speed. Score distillation accelerates diffusion sampling but with performance degradation, while post-training with direct policy optimization (DPO) boosts performance using preference data. This paper proposes Distillation-DPO, a novel diffusion distillation framework for LiDAR scene completion with preference aligment. First, the student model generates paired completion scenes with different initial noises. Second, using LiDAR scene evaluation metrics as preference, we construct winning and losing sample pairs. Such construction is reasonable, since most LiDAR scene metrics are informative but non-differentiable to be optimized directly. Third, Distillation-DPO optimizes the student model by exploiting the difference in score functions between the teacher and student models on the paired completion scenes. Such procedure is repeated until convergence. Extensive experiments demonstrate that, compared to state-of-the-art LiDAR scene completion diffusion models, Distillation-DPO achieves higher-quality scene completion while accelerating the completion speed by more than 5-fold. Our method is the first to explore adopting preference learning in distillation to the best of our knowledge and provide insights into preference-aligned distillation. Our code is public available on https://github.com/happyw1nd/DistillationDPO. |
| title | Diffusion Distillation With Direct Preference Optimization For Efficient 3D LiDAR Scene Completion |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2504.11447 |