Delve into art and creativity using Python

Transcript

1. Delve Into Art and Creativity with Python Aparna Krishnakumar Next Tech Lab SRM University
2. Next Tech Lab SRM University
3. Deep Learning what has been done and what can be done
4. Culture/ Application Deep Learning Architecture Creativity Basic Advanced ‘ART’ Culture/ Application Libraries Art
5. Why Python? • Easy implementation • Powerful, Image processing libraries • comprehensive material • Open Source • AI Frameworks- TensorFlow, Keras, Pytorch • Community • Research Oriented
6. Artificial Intelligence-Deep Learning
7. Training examples X features Data
8. Neural Networks Weights after Activation function back propagation Error cost function Sigmoid function a=g(w*x+b)
10. Height X width X channel pixel values :0-255
11. Convolutional Neural Networks Training examples X features
12. 1. Convolution 2. Max Pooling 3.RELU 4. Fully connected Layer
13. Conv1 Conv2 Conv3 Conv4 more detail
14. Tate Britain AI exhibit Recognition 2016 Tate Reuters
15. Style Transfer Leon A. Gatys, Alexander S. Ecker, Matthias Bethge
16. Image1 Style Content Noise Using weights that minimise the cost function Output Conv 4 Max Pooling Alpa, beta Gram Matrix Style on style genetic algorithms Conv 2? other types of pooling? Experiment
17. Style Transfer + Culture
18. Deep Greetings
19. Deep Dream Features Snapshot of inner workings feed as input Iterate over Outputs: Hybrid organisms Fuse cultures Scans/ medical Imaginary Experiment
20. Foster The People Doing it for the Money
21. GANS Discriminator Generator Training Data
24. Deep Rangoli
25. Coded Couture • Abstract • Style Transfer • Creative Coding • GANS • Augmented Intelligence
26. Why should this be explored further? • Evolution, expression • Innovation • applications • Architectures • logic • socio-economic applications
27. @Aparnaakk aparnakrishnakumar.com @Aparnaakk “ there’s more than one way to do it” (Talk slides and the code for mentioned elements can be found here) Feel free to reach out! aparnaakk@gmail.com