Notice
Transfer Learning and Visualization of Neural Networks for Artistic Images
- document 1 document 2 document 3
- niveau 1 niveau 2 niveau 3
Descriptif
Transfer learning from large-scale natural image datasets, particularly ImageNet, fine-tuning standard deep convolutional neural network models and using the corresponding pre-trained network have become the de facto method for art analysis applications. Nevertheless, there are large differences in dataset sizes, image style and task specifications between natural image classification and the target artistic images, and there is little understanding of the effects of transfer learning. In this work, we explore some properties of transfer learning for artistic images. We compared different ways to obtain an image classifier: fine-tuning, or not, of pre-trained models and training models from scratch. We also use feature visualization techniques in order to understand more precisely what the network learned on those specific artistic datasets. Those visualization of deep neural networks internal representations can help to highlight how neural networks build up their "understanding" of images. We observed that the network could specify some pre-trained filters in order to adapt them to the new modality of images. On the other hand, the network can also learn new, highly structured filters specific to artistic images when the lower-level layers of the initial model are "frozen". In particular, it is possible to obtain classifiers with equivalent classification performances but with different hidden representations, that can be specific to artistic images or not.
Nicolas Gonthier, Yann Gousseau & Saïd Ladjal (Télécom Paris)
Thème
Dans la même collection
-
Angelic Movement. Exploring and Understanding Art, Iconography and Composition with Machine Learning
2e intervention de la 2e session du colloque DHNord 2020
-
ImageGraph: A Visual Programming Language for the Visual Digital Humanities
1ère intervention de la 5e session de DHNord 2020
-
Tracking the Circulation of Images Digitally: From Artistic Cartography to the Study of Visual Cont…
JOYEUX-PRUNEL Béatrice
1ère intervention de la 1ère session du colloque DHNord 2020
-
From Artificial Intelligence to Aesthetic Intelligence
Table ronde de la 3 journée du colloque DHNord 2020
-
Paintings by Artificial Intelligence. Large-scale search for visual similarities
4e intervention de la 1ère session du colloque DHNord 2020
-
Une machine qui lit les images : une histoire socio-technique de l'intelligence artificielle en vis…
CARDON Dominique
3e intervention de la 3e session du colloque DHNord 2020
-
"I Want to Dance": Comparative K-Pop Choreography Analysis Through Deep-Learning Pose Estimation
3e intervention de la 2e session du colloque DHNord 2020
-
Distant Viewing Toolkit: Software for Analysing Visual Culture
2e intervention de la 5e session du colloque DHNord 2020
-
The Art Historical Image Collection at Columbia University: Automating Research on Its Construction…
2e intervention de la 1ère session du colloque DHNord 2020
-
The Image-Theories behind Computer Vision
1ère intervention de la 3e session du colloque DHNord 2020
-
Understanding Images Through Machine Learning and Deep Learning
Table ronde de la 2e journée du colloque DHNord 2020
-
The Augmented Dataset: Artistic Appropriations of Gans and Their Bearings on Ethical Considerations…
Intervention de la 4e session du colloque DHNord 2020