I can make a detailed description of an animal cell, provide examples of illustrations. But I can’t get any satisfactory illustration with organelle labels from GPT or any other model. The best results of generation are attached. Are there any ways to get the desired illustration?
Frontier models struggle with even a small amount of text, this is asking too much currently. Perhaps you could fix the labels in post, but even then are you sure it’s not hallucinating some anatomy in any case? Image production is great for creative intent, but scientific illustration?
Thank you. Yes, I understand that the task is even more difficult than it may seem at first glance. There are a huge number of examples of animal cell illustrations. And all of them have very little to do with electronic photographs of real cells. That is, learning from illustrations is a dubious idea in itself. Perhaps a more correct path would be to create illustrations based on actual photographs and apply captions to them. Illustrations are good because they allow you to remove unnecessary noise from photographs and highlight essential points. But so far I do not understand how to even approach such a task using models.
I really like this idea