I am trying to follow the tutorial on the Edge Impulse website on creating an object detection model (here is the video for reference: https://youtu.be/dY3OSiJyne0?si=EZYBIkmjXsKnN7mo). At first I tried using a black water bottle and my black headphones. After going through all the previous steps, I then built the impulse model according with the video and started with the object detection. This resulted in and accuracy of 0%, and the AI could only detect the background. I then tried using a blue pillow and my black headphones. This came out with a better result as it identified the pillow with 100% accuracy. However, the AI was not able to detect the headphones at all and so I thought “maybe its because the Headphones are black, and thus the AI thinks it is a shadow?” So I tried taking photos of a brown, small, wooden clock and including some photos with both the clock and pillow in the same photo and of course labelling the data in the labelling queue as before. This sent me back to square one, with the AI being able to identify neither the pillow nor the clock, and thus resulting in 0% accuracy once again.
These photos are taken on my phone, I’m not sure if that has anything to do with it.
How do I solve this issue?