250 points by nerd2year 5 months ago flag hide 19 comments
johnsmith 5 months ago next
This is such an interesting topic! I've been messing around with generative art and neural networks myself, can't wait to read this article.
janedoe 5 months ago next
I've been curious about the intersection of art and AI for a while, subscribed to this thread to learn more. Excited to see all the comments!
pixelartking 5 months ago next
There are a few great game dev communities playing around with GANs for in-game content generation. Really interested in exploring the creative side of AI.
coasttocoast 5 months ago next
I recently attempted to build my own AI-driven short stories generator and I was amazed with the results. Now I am even more curious about art generation.
aimastermind 5 months ago next
"Palimpsest: AI Text Generation" is a very interesting article about text generation through machine learning. You might find it intriguing in your journey.
mlmastermind 5 months ago next
@aimastermind Thank you for recommending the Palimpsest article! I found it very clear and helpful for understanding text generation using machine learning methods. For anyone looking for an introduction to this subject, this would be a great starting point.
technophile 5 months ago prev next
Just started working with TensorFlow and this post couldn't have come at a better time. I'm hoping this article can provide some insight on how to approach generative art with neural networks.
datasciencerookie 5 months ago next
Same here! Started with some Keras tutorials for image classification but never thought about using it for art generation. I find the concept mind blowing.
deeplearningart 5 months ago next
I definitely recommend checking out the work of @sophiacaptain, she has some amazing art generated from neural networks.
sophiacaptain 5 months ago next
@deeplearningart thank you for the shout-out! To answer some questions, I use a mix of generative adversarial networks (GANs) and variational auto-encoders (VAEs).
deeplearningart 5 months ago next
@sophiacaptain Do you have any suggestions for resources that would help a beginner even further understand the underlying concepts and get started with GANs and VAEs for art generation (video courses, books, etc.)?
sophiacaptain 5 months ago next
@deeplearningart One resource I think is great for learning about GANs and VAEs is the GANs course on Coursera by Imperial College London. I love Christos Sakaridis, the instructor, and enjoyed the machine learning course from Imperial College that he teaches as well. ๐
alexjoestar 5 months ago prev next
@johnsmith hey John, if you share your personal experience I would love to read about it! It's always better getting insights from people actually trying things out.
thegetgo 5 months ago next
@pythonenthusiast just shared a GitHub repo where I found exactly that, great minds think alike! I'll give it a read.
julielovecode 5 months ago prev next
@thegetgo I checked out that repo, it looks great! I'm excited to learn more about AI-generated art.
helenaspiring 5 months ago prev next
@alexjoestar I've read some tutorials and watched videos about generative art through TensorFlow, and I've started recreating some of the examples I've found online. However I am still a beginner, so I don't have any significant projects to share yet. ๐ค
pixelartking 5 months ago next
Hey @helenaspiring, if you are looking for datasets, check out Kaggle's generative art community, they usually have datasets provided for artistic exploration and generative experiments. Good luck with your learning journey!
aiwizard 5 months ago prev next
From my understanding, there are some challenges in acquiring quality datasets to train generative models, even more so with artistic ones. Is that something you've faced with your projects?
technophile 5 months ago next
@aiwizard Yes, that's a common challenge. I usually try to find Creative Commons-licensed images on platforms like Flickr to help me with this obstacle. Also, I've experimented with unsupervised learning methods as they do not entirely depend on having large datasets. It would be interesting to know whether others have found other ways around it.