Datasets:
The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.
FFHQ-60K Numpy — Processed Face Dataset
A lightweight version of the FFHQ (Flickr-Faces-HQ) dataset containing 60,000 high-quality face images.
This version provides NumPy (.npy) arrays for fast loading and training in deep learning frameworks such as PyTorch and TensorFlow.
Dataset Summary
This dataset is a preprocessed subset of the original FFHQ (70,000 images) dataset.
All images have been:
- Resized to 128×128 pixels ,
- Converted to NumPy arrays (float32),
- Normalized to
[-1,1]range.
This version significantly speeds up training for models like StyleGAN, Diffusion Models, or Autoencoders.
Dataset Structure
| Component | Description |
|---|---|
images.npy |
NumPy array of shape (60000, 128, 128, 3) |
README.md |
This documentation file |
Split: No predefined train/test split.
- Downloads last month
- 10