I have an image that is of size 3.8MB and of resolution 6000x4000 (I have also tried scaling to the suggested resolution)
http://bbarker.co.uk/tagging/UTILS/tmpJjD6KG.jpg
If I pass it to the vision system via the demo page here:
https://cloud.google.com/vision/
It detects the the text within fine
If I pass it to the system using the HTTP API, or the dashboard sandbox, for this image it reliably returns:
{u'responses': [{}]}
Whilst other images using the same code work as expected. I could reduce the resolution, but I'm trying to understand what is different about this image from others, and why it works via the demo page but not the HTTP API. Any ideas? Is it possible the demo page processes the image in some way rather than simply passing it directly to the API...?