supersambat/faceupdat-4x-upscaler

4x face photo upscaler trained on FaceUp dataset using DAT architecture. Improved version of FFHQDAT.

Public
8 runs
License
Weights
Paper
Playground API Examples README Versions

4x-FaceUpDAT Face Upscaler

A 4x super-resolution model specifically designed for face and portrait photos. Trained on the FaceUp dataset using the DAT (Dual Aggregation Transformer) architecture. This is an improved version of the earlier FFHQDAT model, with better detail preservation and fewer artifacts.

Model Details

  • Architecture: DAT (Dual Aggregation Transformer)
  • Scale: 4x upscaling
  • Training data: FaceUp (10,000 curated face images)
  • Training iterations: 140,000
  • Best for: Portrait and face photo enhancement

Usage

Upload a face/portrait image and the model will return a 4x upscaled version with enhanced facial details.

Credits & Acknowledgments

This model was originally created by Helaman and published on OpenModelDB. Full credit goes to the original author for training and releasing this model.

This Replicate deployment is a wrapper for convenient API access. The original model is licensed under CC-BY-4.0.

  • 4x-FFHQDAT — Earlier version trained on FFHQ dataset
Model created