gszfwsb/NCFM
PyTorch implementation of a CVPR 2025 paper that distills large image datasets into compact synthetic datasets using neural characteristic function matching.
Not currently ranked — collecting fresh signals.
star history
This repository provides the official implementation of NCFM (Neural Characteristic Function Matching), a dataset distillation method published at CVPR 2025. The method synthesizes small distilled datasets that can effectively train models, significantly reducing data requirements while maintaining performance. It uses neural characteristic functions and minmax optimization to match the statistical properties of the original dataset, applicable to standard benchmarks like CIFAR-10/100 and ImageNet.
Frequently asked
- What is gszfwsb/NCFM?
- PyTorch implementation of a CVPR 2025 paper that distills large image datasets into compact synthetic datasets using neural characteristic function matching.
- Is NCFM open source?
- Yes — gszfwsb/NCFM is an open-source project tracked on heatdrop.
- What language is NCFM written in?
- gszfwsb/NCFM is primarily written in Python.
- How popular is NCFM?
- gszfwsb/NCFM has 414 stars on GitHub.
- Where can I find NCFM?
- gszfwsb/NCFM is on GitHub at https://github.com/gszfwsb/NCFM.