work
Visualize What AI Learns (Series)
| category | Digital art |
| subject | Nature, Human figure, Abstract, Animal |
| tags | IA, Artificial Intelligence, AI, Intelligenza Artificiale |
| base | 60 cm |
| height | 60 cm |
| depth | 5 cm |
| year | 2021 |
Serie: Visualize What AI Learns /
Inception V3 - Mixed6d_concat - channel-21
Inception V3 - Mixed6d_concat - channel-12
Inception V3 - Mixed6d_concat - channel-0
Inception 1 - conv2d2 - channel-24
ResNet 50 V2 - block2-unit_3-add-channel-52
60x60 cm, 2021
Digital image NFT, Print: digigraphie on metallic paper
The images in this series do not represent physical objects, human artifacts and are not even patterns generated by an artificial intelligence. These, in fact, are actual photographs of the mind of an artificial intelligence. They show what it has learned, how it perceives and extrapolates meaning from the world. Halfway between scientific photography and conceptual art, the images in this series reproduce abstract models that in the case of the human mind we would call, taking up a philosophical term, categories. Technically, these are visualizations of functions derived from automatic learning of convolutional neural networks. Convolutional neural networks are used in the field of Computer Vision, the science that deals with methods and algorithms to analyze images and videos in order to extract information and meanings. In this field, convolutional neural networks (which simulate the functioning of the human primary visual cortex) are among the most important and advanced techniques.
The graphic visualization was obtained through a Python code that reconstructed the images starting from some of the most important networks used by the community of developers and researchers.
Inception V3 - Mixed6d_concat - channel-21
Inception V3 - Mixed6d_concat - channel-12
Inception V3 - Mixed6d_concat - channel-0
Inception 1 - conv2d2 - channel-24
ResNet 50 V2 - block2-unit_3-add-channel-52
60x60 cm, 2021
Digital image NFT, Print: digigraphie on metallic paper
The images in this series do not represent physical objects, human artifacts and are not even patterns generated by an artificial intelligence. These, in fact, are actual photographs of the mind of an artificial intelligence. They show what it has learned, how it perceives and extrapolates meaning from the world. Halfway between scientific photography and conceptual art, the images in this series reproduce abstract models that in the case of the human mind we would call, taking up a philosophical term, categories. Technically, these are visualizations of functions derived from automatic learning of convolutional neural networks. Convolutional neural networks are used in the field of Computer Vision, the science that deals with methods and algorithms to analyze images and videos in order to extract information and meanings. In this field, convolutional neural networks (which simulate the functioning of the human primary visual cortex) are among the most important and advanced techniques.
The graphic visualization was obtained through a Python code that reconstructed the images starting from some of the most important networks used by the community of developers and researchers.
