Dataset Color Quantization: A Training-Oriented Framework for Dataset-Level Compression
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910036491501568 |
|---|---|
| author | Yu, Chenyue Xiao, Lingao Deng, Jinhong Tsang, Ivor W. He, Yang |
| author_facet | Yu, Chenyue Xiao, Lingao Deng, Jinhong Tsang, Ivor W. He, Yang |
| contents | Large-scale image datasets are fundamental to deep learning, but their high storage demands pose challenges for deployment in resource-constrained environments. While existing approaches reduce dataset size by discarding samples, they often ignore the significant redundancy within each image -- particularly in the color space. To address this, we propose Dataset Color Quantization (DCQ), a unified framework that compresses visual datasets by reducing color-space redundancy while preserving information crucial for model training. DCQ achieves this by enforcing consistent palette representations across similar images, selectively retaining semantically important colors guided by model perception, and maintaining structural details necessary for effective feature learning. Extensive experiments across CIFAR-10, CIFAR-100, Tiny-ImageNet, and ImageNet-1K show that DCQ significantly improves training performance under aggressive compression, offering a scalable and robust solution for dataset-level storage reduction. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_20650 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Dataset Color Quantization: A Training-Oriented Framework for Dataset-Level Compression Yu, Chenyue Xiao, Lingao Deng, Jinhong Tsang, Ivor W. He, Yang Computer Vision and Pattern Recognition Artificial Intelligence Large-scale image datasets are fundamental to deep learning, but their high storage demands pose challenges for deployment in resource-constrained environments. While existing approaches reduce dataset size by discarding samples, they often ignore the significant redundancy within each image -- particularly in the color space. To address this, we propose Dataset Color Quantization (DCQ), a unified framework that compresses visual datasets by reducing color-space redundancy while preserving information crucial for model training. DCQ achieves this by enforcing consistent palette representations across similar images, selectively retaining semantically important colors guided by model perception, and maintaining structural details necessary for effective feature learning. Extensive experiments across CIFAR-10, CIFAR-100, Tiny-ImageNet, and ImageNet-1K show that DCQ significantly improves training performance under aggressive compression, offering a scalable and robust solution for dataset-level storage reduction. |
| title | Dataset Color Quantization: A Training-Oriented Framework for Dataset-Level Compression |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2602.20650 |