I am working on a project where part of it is to detect PV module arrays, I trained few object detection models through TensorFlow Object Detection API and the problem I got is that the trained models do detect narrow arrays very well, but do not work that well on wider objects: enter image description here enter image description here

On the second picture model caught 3 objects but with low probability, and 2 are still missing. I tried following models already:

  • ssd mobilenet v2 640x640
  • ssd mobilenet v2 320x320
  • efficient det4
  • ssd resnet 50 1024x1024

All of them have the same problem with wider objects. I might also add that in the training data there were examples of wider objects too, it's hard to tell, but maybe even more than narrow ones.

What can be the reason of this? Is there anyway to increase accuracy of detection for wider objects? Can some particular model or architecture be better for this application (maybe one of the YOLOs)?

EDIT: I might add, that so far I use 800x800px images, but I can change it.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.