I just got access to ChatGPT being able to view my images. So, I put it to a very simple test on a diagram I'm revising for general usage, see attached image. The response definitely captures all of the services being utilized in the diagram, see attached image.
Finally, i asked how security could be improved, to which it gave some great suggestions. I'm definitely curious to how other computer vision services stack up to this. It's a very simple use case, but it can save a lot of time to get you some quick recommendations to start building on top of.
UPDATE: There's a model based on the Llama series that has similar functionality, but is definitely lacking on first pass. I passed it the same picture and prompt and it wasn't able to get the services as correct as ChatGPT-V. Although, i'm sure it will get better and to be fair, it gives you an easy way to train your own model up from their base, i'm grateful it exists. P.S. I would have shared the actual conversation, but it's not supported just yet with ChatGPT.
Which computer vision models have you all tried thus far?
Let us know!