Hello
A while now, I have been looking for a GenAI model/system/software where, apart from getting the final result image, all the iterations happening in the background can also be viewed.
This is to visually study, practical research, how de iterations evolve from noise to a final image result.
Can any of you point me in the right direction?
Like this?
Hello!
For your use case, where you’d like to see the iterative process behind the generation of an image from noise to the final result, you might want to explore Stable Diffusion models. Many implementations of Stable Diffusion, like the ones in Hugging Face or RunwayML, provide the option to view or log intermediate steps, showing the evolution of the image as it is generated.
You can also experiment with GANs (Generative Adversarial Networks), specifically the Progressive GAN or StyleGAN, which allow you to observe how an image progresses from random noise through each phase of refinement. Some tools also offer the ability to visualize latent space traversal, where you can track how the model’s input evolves into the final output.
Additionally, platforms like Google Colab and Kaggle Kernels may offer notebooks where you can access and manipulate these models to visualize intermediate steps during image generation.
I hope this helps, and best of luck with your research!
@John6666 As soon as I have time I’ll try out the callbacks.
@Alanturner2 Wow, thanks, you describe my question clearly and take me directly in the right direction!
Now figuring out how that works, do you know a tutorial, article, …, where this is explained in more depth or can this be done by the ‘callbacks’?