Hello @MakarovDs777 !
Funny project ! Unfortunately, while explorating some pure noise, the odds of seeing by chance the 3D modeling of an actual object is quite low 
Do you know about GANs ? If not, you should definively have a look !
GAN (Generative Advertial Networks) are the heart behind what became later the Stable Diffusion AIs used by Midjourney, ChatGPT, Grok, to generate imagesâŚ
- One âGeneratorâ starts from noise and generate images
- One âDiscrimatorâ which is trained on real images says how much itâs ânoiseâ.
- The first one edit the noise so that the second one sees a less ânoisyâ image
- Eventually, the two âadvertialâ network converge toward something
- Changing the very first input noise change totally the final converged destination
If you know about Python, you should definively give it a try ! (tensorflow, keras)
Back in 2014 when I started this it was a nightmare, no library existed, papers was partially published, etc⌠Now that we are in the world of OpenSource + AIs ⌠You can go fast 