User:Alexander Roidl/everything: Difference between revisions
No edit summary |
No edit summary |
||
| Line 5: | Line 5: | ||
https://www.youtube.com/watch?v=iGZpJZhqEME&index=19&list=PLd3hlSJsX_Imk_BPmB_H3AQjFKZS9XgZm | https://www.youtube.com/watch?v=iGZpJZhqEME&index=19&list=PLd3hlSJsX_Imk_BPmB_H3AQjFKZS9XgZm | ||
* Things that we can not explain with only words anymore, there is no language for it. | |||
(drone shadows, making it visible) | |||
* How can we describe images? Features in machine learning | |||
* Flusser > image reduced to 0 dimensions > abstraction of reality | |||
=Image Aesthetics= | |||
==Machine Learning Algorithms rating image aesthetics== | ==Machine Learning Algorithms rating image aesthetics== | ||
Paper: Rating Image Aesthetics using Deep Learning by Xin Lu, Zhe Lin, Hailin Jin, Jianchao Yang, and James. Z. Wang | Paper: Rating Image Aesthetics using Deep Learning by Xin Lu, Zhe Lin, Hailin Jin, Jianchao Yang, and James. Z. Wang | ||
| Line 15: | Line 20: | ||
-> how can the algorithm be convinced by art? (new criteria for aesthetic) | -> how can the algorithm be convinced by art? (new criteria for aesthetic) | ||
=Machine Learning as intervention / Augmented Reality= | =Machine Learning as intervention in human vision / Augmented Reality= | ||
(Is it augmented reality if what we see is not real? Maybe it is more intervened reality) | (Is it augmented reality if what we see is not real? Maybe it is more intervened reality) | ||
==ml algorithms in smartphone cameras== | ==ml algorithms in smartphone cameras== | ||
| Line 29: | Line 34: | ||
http://blog.otoro.net/2015/06/19/neural-network-generative-art/ | http://blog.otoro.net/2015/06/19/neural-network-generative-art/ | ||
what happens without training | what happens without training? | ||
→ direct visualisation of the algorithms (the algorithm is the art) | |||
| Line 41: | Line 47: | ||
= | =The computational model= | ||
From the database the computer is being trained. While feeding these datasets into huge models we lose sight of it's connections, that make sense for only the computer now on. A model that tries to describe reality in order to generate, to analyze or predict. | From the database the computer is being trained. While feeding these datasets into huge models we lose sight of it's connections, that make sense for only the computer now on. A model that tries to describe reality in order to generate, to analyze or predict. | ||
| Line 76: | Line 82: | ||
=new aesthetics= | =new aesthetics= | ||
===new visual forms=== | ===new visual forms=== | ||
| Line 122: | Line 125: | ||
Error images as understanding, when you cant see the model anymore. | Error images as understanding, when you cant see the model anymore. | ||
====Changable images==== | ====Changable images==== | ||
(Emojis render differently on every device) | * (Emojis render differently on every device) | ||
* Giving just the information of what it is > image | |||
Giving just the information of what it is > image | * Personalised images through machine learning? | ||
Personalised images through machine learning? | |||
===Infinite production=== | ===Infinite production=== | ||
| Line 141: | Line 137: | ||
=Prototyping ideas= | =Prototyping ideas= | ||
==Every image ever== | |||
constantly changing image that would generate every image ever that is possible with that set of pixels | constantly changing image that would generate every image ever that is possible with that set of pixels | ||
(an image recognition algorithm could run over it and save all images that contain something) | (an image recognition algorithm could run over it and save all images that contain something) | ||
==new earth== | |||
Generate new images from satellite imagery | Generate new images from satellite imagery | ||
==endless production== | |||
producing endless material | producing endless material | ||
(removing it after to free resources) | (removing it after to free resources / endless snapchat) | ||
=Resources= | =Resources= | ||
==Neural Networks== | |||
* http://neuralnetworksanddeeplearning.com/chap1.html | * http://neuralnetworksanddeeplearning.com/chap1.html | ||
* https://www.youtube.com/watch?v=aircAruvnKk | * https://www.youtube.com/watch?v=aircAruvnKk | ||
Revision as of 14:04, 25 October 2018
Sorting ideas out
The Language of Images
Bag of Features
https://www.youtube.com/watch?v=iGZpJZhqEME&index=19&list=PLd3hlSJsX_Imk_BPmB_H3AQjFKZS9XgZm
- Things that we can not explain with only words anymore, there is no language for it.
(drone shadows, making it visible)
- How can we describe images? Features in machine learning
- Flusser > image reduced to 0 dimensions > abstraction of reality
Image Aesthetics
Machine Learning Algorithms rating image aesthetics
Paper: Rating Image Aesthetics using Deep Learning by Xin Lu, Zhe Lin, Hailin Jin, Jianchao Yang, and James. Z. Wang
So what does aesthetic mean for an algorithm? Is it just categorisation? Therefore: what is in the dataset? -> making art that pleases the machine -> how can the algorithm be convinced by art? (new criteria for aesthetic)
Machine Learning as intervention in human vision / Augmented Reality
(Is it augmented reality if what we see is not real? Maybe it is more intervened reality)
ml algorithms in smartphone cameras
- real time intervention in reality
- Overlapping of recognition (faces / boxes)
Images Databases
Image Net
Art with algorithms
ml art
http://blog.otoro.net/2015/06/19/neural-network-generative-art/
what happens without training? → direct visualisation of the algorithms (the algorithm is the art)
Computer Vision
Irritating / Failing AI
Stock image platforms
also interesting with their stamp on it
The computational model
From the database the computer is being trained. While feeding these datasets into huge models we lose sight of it's connections, that make sense for only the computer now on. A model that tries to describe reality in order to generate, to analyze or predict.
I want to put a special focus on generative models and discriminative models of the world.
While these models simplify reality by trying to calculate probabilities and reduce to features, in the same time it makes reality more complicated. These computed models are black boxes. Mostly it is even impossible to see the data it is being trained on.
From features to reduced reality
So an image is being reduced to its most contrasting points, to its pixels that hold a certain array of color values. But what is it that makes an image? If I was to describe an image, I wouln't say: Oh, there is some contrast going on in the left corner, lots of brightness in the middle and
A sentence is being reduced to its words and connections. But can we describe the value of a sentence by only this features?
Training against myself
It becomes even weirder when we look at algorithms that create models by learning from themselves. So we do not only encounter the problem of creating an abstract model, even the database, that otherwise enables us to gather insights on why models act how they act, is incomplete.
How models are built
- model is built by computer scientists
- often incomplete / lacking / biased databases
- selecting features > generalizing
unsemantic everything
Semantic sorting can be filtered trough models > Unsemantic web
Models turn chaos into sense (for a machine)
Bending the model - an experimental approach in understanding
How can we understand machine learning models by their mistakes and missunderstandings
> what can we learn from that different view on reality?
How can we bring models to fail or to its limits, how can you abuse them?
new aesthetics
new visual forms
computer vision
How visual material becomes more important in order to understand these technologies
Errors in images > what do they mean?
Image Manipulation: Small things that we cannot see, but computers can. (Clones image zones)
https://theblog.adobe.com/spotting-image-manipulation-ai/
Invisible for humans
Main question: what makes these aesthetics so different from
What is new and what is old?
New Algorithmic Form
How does machine learning change the generation of form and image. What are the implications on culture and art production through those generated forms. Side question: what makes these new aesthetics so different from the old?
Art Historical Context
Relation to Production of Art and generation through algorithm. (Walter Benjamin -> from mechanical reproduction to digital production. Art is not only reproduced by machines but it is purely produced by them)
On Computer Vision
Visibility
What is visible / invisible for machines / humans.
- Image Manipulation: Small things that we cannot see, but computers can. (Clones image zones) https://theblog.adobe.com/spotting-image-manipulation-ai/
- Satellite Target images
- Weather satellite imagery
Making visible as form of understanding.
- images with bounding boxes
Error images as understanding, when you cant see the model anymore.
Changable images
- (Emojis render differently on every device)
- Giving just the information of what it is > image
- Personalised images through machine learning?
Infinite production
Prototyping ideas
Every image ever
constantly changing image that would generate every image ever that is possible with that set of pixels
(an image recognition algorithm could run over it and save all images that contain something)
new earth
Generate new images from satellite imagery
endless production
producing endless material (removing it after to free resources / endless snapchat)