Currently chatGPT Vision has a problem of not finding the pixel coordinates of a object in the image. To address that problem, I've made a sectioning tool that repeatedly divides the image into 9 parts to then find the pixel coordinates of those objects albeit with poor accuracy.
Show HN: Making ChatGPT Vision output pixel coordinates of objects | Heykuki News