JPEG AIC-3 Dataset

This page describes the test materials that may be used to evaluate the contributions to the JPEG AIC CfC on Subjective Image Quality Assessment [1], namely the JPEG AIC-3 Dataset.

Contributors to the JPEG AIC CfC may use the JPEG AIC-3 dataset to evaluate the proposed subjective quality assessment contribution or may use a different dataset, which must be included as part of the contribution.

More information on the encoding of the images and on the selection of the distorted images is available in [2]. Please refer to [3] for more detailed information.


The JPEG AIC-3 dataset comprises 10 uncompressed original images of different resolutions and contents: object, human portrait, food, computer-generated image, animal, a scene with water, a night scene, fabric/fine texture, landscape, and buildings.

Distorted images

The dataset includes images distorted using the following codecs:

  • JPEG
  • JPEG 2000
  • HEVC Intra
  • VVC Intra
  • AVIF


The dataset can be downloaded through FTP by using dedicated FTP clients, such as FileZilla or FireFTP (we recommend using FileZilla):

Protocol: FTP
FTP address:
Username: [email protected]
Password: .L:p*O
FTP port: 21
Folder: 2023-01

After you connect, choose from the remote site the folder “JPEG AIC-3 Dataset” and then “2023-01“. The original images can be found in the folder “original“, while the distorted images can be found in the folder “decoded“. The total size of the dataset is ~1.5 GB.

The decoded images are organized into 10 different sub-folders, one for each content. The naming convention is the following:


where <QUALITY> is a number between 1 and 10, where 1 indicates the best quality and 10 worst quality. More information can be found in the file “info.csv” include in the “decoded” subfolder.


The data are licensed under a Creative Commons CC0 (

Conditions of use

In no event shall the Ecole Polytechnique Fédérale de Lausanne (EPFL) be liable to any party for direct, indirect, special, incidental, or consequential damages arising out of the use of the data and its documentation. The Ecole Polytechnique Fédérale de Lausanne (EPFL) specifically disclaims any warranties. The data provided hereunder is on an “as is” basis and the Ecole Polytechnique Fédérale de Lausanne (EPFL) has no obligation to provide maintenance, support, updates, enhancements, or modifications.

If you wish to use any of the provided material in your research, we kindly ask you to cite [3].


[1] ISO/IEC JTC1/SC29/WG1 N100311, “Final Call for Contributions on Subjective Image Quality Assessment”, Online, October 2022. [Access:]

[2] ISO/IEC JTC1/SC29/WG1 N100334, “Common Test Conditions on Subjective Image Quality Assessment”, Online, October 2022. [Access:]

[3] Michela Testolina, Vlad Hosu, Mohsen Jenadeleh, Davi Lazzarotto, Dietmar Saupe, and Touradj Ebrahimi. “JPEG AIC-3 Dataset: Towards Defining the High Quality to Nearly Visually Lossless Quality Range.” 2023 15th International Conference on Quality of Multimedia Experience (QoMEX). IEEE, 2023.


In case of any questions or problems, please contact Michela Testolina at [email protected].