Painting Authentication Using CNNs and Sliding Window Feature Extraction
Last update: December - 2025
Painting authentication is a complex task, relying on a combination of connoisseurship and technical analysis.
This study focuses on the authentication of a single painting attributed to Paolo Veronese,
using a convolutional neural network (CNN) approach tailored to extreme data scarcity.
You can download the MATLAB source code by clicking here (cnnPainting.zip).
The dataset for training and testing the CNN can be downloaded by clicking here (dataset.zip, 19 MB).
Once unzipped the file cnnPainting.zip, you will find the following files:
- CNN_validation.m: MATLAB file for validating the CNN
- CNN_test.m: MATLAB file for testing the CNN
- MobileNetV2_validation.m: MATLAB file for validating with MobileNetV2
- MobileNetV2_test.m: MATLAB file for testing with MobileNetV2
- extractPatches.m, prepareDataCNN.m and heatMapProbVeronese.m:
complementary MATLAB files for sliding-window computation and generation of probability heatmaps
In the folder with the MATLAB code, unzip the dataset.zip file into a subfolder dataset/,
what creates the folders dataset/non-veronese/, dataset/test/ and
dataset/veronese/. These folders contain the BMP files with all the images to train and test the CNN.
REQUERIMENTS for execution
Software:
CNN/MobileNetV2 validation
From the MATLAB prompt, run the files CNN_validation.m and MobileNetV2_validation.m. Performance metrics will be
shown at the MATLAB console, and the figure with the ROC-curve will be displayed.
CNN/MobileNetV2 test
From the MATLAB prompt, run the files CNN_test.m and MobileNetV2_test.m. Average Veronese probability for each test
painting will be shown at the MATLAB console, and the figures with the Veronese probability heatmap for
each test painting will be displayed.
CITATION
"Painting Authentication Using CNNs and Sliding Window Feature Extraction"
Juan Ruiz de Miras, José Luis Víchez, María José Gacto and Domingo Martín
Frontiers in Artificial Intelligence 8, 2026
https://doi.org/10.3389/frai.2025.1738444
CONTACT AND SUGGESTIONS
For contact information click here
FUNDING
This software is part of the I+D+i project PID2024-161348OB-I00 granted by the Spanish Ministry of Science, Innovation and University
