Alexander Mordvintsev
Just before DeepDream: 1000 classes #3, 2015/01
Fine Art Print on paper
Size: 60 × 60 cm
The work is also available as a unique NFT (minted upon request) for 5 ETH
Description:
The Deep Dream algorithm was created in a series of experiments that led to the generation of psychedelic images full of intricate details that swept across the globe in the summer of 2015. Slides 4-8 show images that were created on the way to the final algorithm. Scientific progress is enabled by accumulation of ideas and building on the work of others. For example, p.4 “Just before DeepDream 1000 classes” shows a few images out of 1000 that were generated by the algorithm inspired by the paper . The network was also different: these images were produced by the AlexNet , created in 2012, which is often cited as one of the main triggering points of the Deep Learning revolution. These images lack the resolution and rich details, but already show promising capabilities of deep nets for image generation. The subsequent slides show steps that Alexander Mordivntsev took to improve the fidelity of the machine generations and finally giving the network more “creative freedom” by formulating the objective for amplification patterns already perceived by the network in a particular input image.
For more information please email us on info@katevassgalerie.com
Exhibited at Untitled Miami Art Fair, Miami, Kate Vass Galerie Booth A43, December 2024
Exhibited at Automat und Mensch 2.0, Kate Vass Galerie, May 2024
Just before DeepDream: 1000 classes #3, 2015/01
Fine Art Print on paper
Size: 60 × 60 cm
The work is also available as a unique NFT (minted upon request) for 5 ETH
Description:
The Deep Dream algorithm was created in a series of experiments that led to the generation of psychedelic images full of intricate details that swept across the globe in the summer of 2015. Slides 4-8 show images that were created on the way to the final algorithm. Scientific progress is enabled by accumulation of ideas and building on the work of others. For example, p.4 “Just before DeepDream 1000 classes” shows a few images out of 1000 that were generated by the algorithm inspired by the paper . The network was also different: these images were produced by the AlexNet , created in 2012, which is often cited as one of the main triggering points of the Deep Learning revolution. These images lack the resolution and rich details, but already show promising capabilities of deep nets for image generation. The subsequent slides show steps that Alexander Mordivntsev took to improve the fidelity of the machine generations and finally giving the network more “creative freedom” by formulating the objective for amplification patterns already perceived by the network in a particular input image.
For more information please email us on info@katevassgalerie.com
Exhibited at Untitled Miami Art Fair, Miami, Kate Vass Galerie Booth A43, December 2024
Exhibited at Automat und Mensch 2.0, Kate Vass Galerie, May 2024
Just before DeepDream: 1000 classes #3, 2015/01
Fine Art Print on paper
Size: 60 × 60 cm
The work is also available as a unique NFT (minted upon request) for 5 ETH
Description:
The Deep Dream algorithm was created in a series of experiments that led to the generation of psychedelic images full of intricate details that swept across the globe in the summer of 2015. Slides 4-8 show images that were created on the way to the final algorithm. Scientific progress is enabled by accumulation of ideas and building on the work of others. For example, p.4 “Just before DeepDream 1000 classes” shows a few images out of 1000 that were generated by the algorithm inspired by the paper . The network was also different: these images were produced by the AlexNet , created in 2012, which is often cited as one of the main triggering points of the Deep Learning revolution. These images lack the resolution and rich details, but already show promising capabilities of deep nets for image generation. The subsequent slides show steps that Alexander Mordivntsev took to improve the fidelity of the machine generations and finally giving the network more “creative freedom” by formulating the objective for amplification patterns already perceived by the network in a particular input image.
For more information please email us on info@katevassgalerie.com
Exhibited at Untitled Miami Art Fair, Miami, Kate Vass Galerie Booth A43, December 2024
Exhibited at Automat und Mensch 2.0, Kate Vass Galerie, May 2024