detect_text cannot find obvious texts

I took a screenshot and ran this example and the detect_text API does detect the text "VINCENNES / PRIX ", though with a lower confidence in the 60-70 range (depending on how I sliced the image) with the "/" and to much higher score without the "/". For this reason, at least in the demo on the AWS console , you will not see a bounding box plotted. Based on this I would agree with your assumption that the "/" may be causing the lower confidence score. I would recommend the following:

1. Re-run this example and observe the API response for the text. If the response does not contain the text at all, please share it here.
2. In your program logic, process/consider lower ranges of confidence scores if the DetectedText contains a "/" (to draw a bounding box etc).