Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2403.06741 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866929374142398464 |
|---|---|
| author | Zhu, Haowei Yang, Ling Yong, Jun-Hai Yin, Hongzhi Jiang, Jiawei Xiao, Meng Zhang, Wentao Wang, Bin |
| author_facet | Zhu, Haowei Yang, Ling Yong, Jun-Hai Yin, Hongzhi Jiang, Jiawei Xiao, Meng Zhang, Wentao Wang, Bin |
| contents | The scale and quality of a dataset significantly impact the performance of deep models. However, acquiring large-scale annotated datasets is both a costly and time-consuming endeavor. To address this challenge, dataset expansion technologies aim to automatically augment datasets, unlocking the full potential of deep models. Current data expansion techniques include image transformation and image synthesis methods. Transformation-based methods introduce only local variations, leading to limited diversity. In contrast, synthesis-based methods generate entirely new content, greatly enhancing informativeness. However, existing synthesis methods carry the risk of distribution deviations, potentially degrading model performance with out-of-distribution samples. In this paper, we propose DistDiff, a training-free data expansion framework based on the distribution-aware diffusion model. DistDiff constructs hierarchical prototypes to approximate the real data distribution, optimizing latent data points within diffusion models with hierarchical energy guidance. We demonstrate its capability to generate distribution-consistent samples, significantly improving data expansion tasks. DistDiff consistently enhances accuracy across a diverse range of datasets compared to models trained solely on original data. Furthermore, our approach consistently outperforms existing synthesis-based techniques and demonstrates compatibility with widely adopted transformation-based augmentation methods. Additionally, the expanded dataset exhibits robustness across various architectural frameworks. Our code is available at https://github.com/haoweiz23/DistDiff |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_06741 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Distribution-Aware Data Expansion with Diffusion Models Zhu, Haowei Yang, Ling Yong, Jun-Hai Yin, Hongzhi Jiang, Jiawei Xiao, Meng Zhang, Wentao Wang, Bin Computer Vision and Pattern Recognition The scale and quality of a dataset significantly impact the performance of deep models. However, acquiring large-scale annotated datasets is both a costly and time-consuming endeavor. To address this challenge, dataset expansion technologies aim to automatically augment datasets, unlocking the full potential of deep models. Current data expansion techniques include image transformation and image synthesis methods. Transformation-based methods introduce only local variations, leading to limited diversity. In contrast, synthesis-based methods generate entirely new content, greatly enhancing informativeness. However, existing synthesis methods carry the risk of distribution deviations, potentially degrading model performance with out-of-distribution samples. In this paper, we propose DistDiff, a training-free data expansion framework based on the distribution-aware diffusion model. DistDiff constructs hierarchical prototypes to approximate the real data distribution, optimizing latent data points within diffusion models with hierarchical energy guidance. We demonstrate its capability to generate distribution-consistent samples, significantly improving data expansion tasks. DistDiff consistently enhances accuracy across a diverse range of datasets compared to models trained solely on original data. Furthermore, our approach consistently outperforms existing synthesis-based techniques and demonstrates compatibility with widely adopted transformation-based augmentation methods. Additionally, the expanded dataset exhibits robustness across various architectural frameworks. Our code is available at https://github.com/haoweiz23/DistDiff |
| title | Distribution-Aware Data Expansion with Diffusion Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2403.06741 |