Update (Jan. 3, 2020): We updated the dataset after noticing a few errors. The following changes were made.

  • Previously, the order of lighting conditions in the readme did not match that of the images. This issue is now fixed.
  • Previously, the order of the captures did not match the order of the provided reconstructions. This issue is now fixed.
  • We removed subject 58 and adjusted the subsequent subject numbers accordingly.
  • Reconstruction code now includes both Python and Matlab functions.
  • We paired up images in test_pairs_subj61-87.txt to provide a standard FCFD face verification test (details in the readme).
  • The readme has been updated according to these changes.

The FlatCam Face Dataset (FCFD) is a dataset containing 23,838 face images of 87 different subjects captured using the FlatCam lensless imaging system. For more information on FlatCam, please click here. The dataset contains images of the faces in a variety of lighting conditions, expressions, angles, and backgrounds. The dataset is available for public and can be accessed via the following Google Drive links:

  • Readme: contains detailed information on the FCFD. [link
  • Raw captures: png files of the raw FlatCam captures [link]
  • Reconstructions: images reconstructed from the FlatCam measurements using Tikhonov least squares reconstruction [link]
  • Webcam: corresponding cropped Webcam images [link]
  • Code: Matlab and Python code to reconstruct images from the raw captures (run demo.py or demo.m for an example) [link]

For more information, please see our paper.

If you use this dataset in your research, please cite the following publication: 

author={J. {Tan} and L. {Niu} and J. K. {Adams} and V. {Boominathan} and J. T. {Robinson} and R. G. {Baraniuk} and A. {Veeraraghavan}}, 
journal={IEEE Trans. Comput. Imag.}, 
title={Face Detection and Verification Using Lensless Cameras},