98 points by artisticdev 6 months ago flag hide 18 comments
artistic_developer 6 months ago next
This is really cool! Love the real-time generative artwork. React and TensorFlow.js integration is on point.
machine_learner 6 months ago next
Thanks! We used TensorFlow.js to train our models quickly and deploy them to the browser with minimal effort.
ai_curious 6 months ago next
What inspired you to work on this project? Did you find any performance limitations or challenges while implementing the generative models in the browser?
artistic_developer 6 months ago next
We loved experimenting with interactive and procedural art and thought integrating real-time ML would be a unique approach. Performance-wise, we made sure to leverage GPU acceleration using WebGL through TensorFlow.js for model inference.
innovative_mind 6 months ago next
Really cool stuff! Collaborative AI art could be an interesting direction to explore with this project. Artists could work together on a single piece of art by sharing control of the ML models. Food for thought.
artistic_developer 6 months ago next
That's true; collaborative AI art would be an exciting extension to our current project! It's definitely in our backlog of ideas to explore. Thank you for suggesting this!
web_creator 6 months ago prev next
Any chance you can share your approach to syncing the generative art with the real-time UI updates? Impressive work!
machine_learner 6 months ago next
Sure! We used React Hooks for state management and a library called 'socket.io' for handling real-time communication. You can check out the source code in the GitHub repo.
performance_buff 6 months ago next
The real-time element makes this app even more interesting. Have you experimented with WebAssembly for potentially speeding up the machine learning processes? I'm assuming you're using WASM with TensorFlow.js behind the scenes.
machine_learner 6 months ago next
Great question! We did explore using WebAssembly but decided to hold off on implementing it for now due to time constraints and browser compatibility issues. Our app mostly relies on WebGL for ML processing, so it's something we might circle back to later.
code_lover 6 months ago prev next
This is such an interesting crossover between art and code! Are there any resources or tutorials you could suggest for beginners interested in combining ML with creative apps?
artistic_developer 6 months ago next
Thank you! We recommend checking out resources like 'Creative Applications of Deep Learning' by Mendel Brott, and exercising your creative coding/ML skills on platforms like openprocessing.org and glitch.com
newbie_coder 6 months ago prev next
I'm new to ML, and this project seems like a great starting point to learn MT in the browser. Would you recommend using TensorFlow.js for my early ML projects, or are there more suitable libraries for beginners?
helpful_hacker 6 months ago next
TensorFlow.js is a great option for beginners as it has a user-friendly API and allows you to leverage existing TensorFlow models. You can also try 'Keras.js', which is a JavaScript library built on top of TensorFlow.js.
future_tech 6 months ago prev next
Have you considered integrating augmented reality (AR) with your generative art? Imagine using this to create art in real-world environments or even to produce on-the-fly visuals for concerts and performances!
vr_enthusiast 6 months ago next
I'd love to see that! For AR integration, I'd suggest looking at libraries like 'A-Frame' or 'Babylon.js'. If you're interested in VR, take a look at 'WebXR', which consists of WebGL-based APIs for AR and VR.
open_source 6 months ago prev next
Did you share the source code for this project on GitHub or another open-source platform? I'd love to take a deeper look and maybe even contribute!
artistic_developer 6 months ago next
Absolutely! You can find the source code and a live demo on our GitHub page. Contributions are more than welcome. Enjoy exploring and playing around with the code!