I talked to a medical professional yesterday, who had been using GenAI images to use in lectures and workshop situations, when teaching medical students to identify tumors etc. The problem with all the major image generation platforms is that they prevent the generation of injuries and abnormalities, so one needs to create prompts that go around these restrictions ("a lump in forearm" instead of "a visible tumor in forearm"), and many kinds of images are not possible (for example, images with blood and injuries).
Are there any image generation services that don't have these kinds of restrictions? Or can you give me pointers on which open-source text-to-image models could be used for creating a customized model for medical education?
This would be hugely beneficial for medical education, since getting the permissions to use this kinds of images is usually very difficult and time-consuming.
[link] [comments]