FOMO not working for two different objects in a single frame

Report bugs
#1

Question/Issue:
Will the FOMO model detect more than one different object in a frame? I have trained the model to detect two different objects (a banana and a cup). After training and retraining the model, when I tried to classify in real-time, with both objects in the frame at the same time, it wasn’t able to detect anything. Is this something I can fix, if so how?

[Provide the project ID]
609080

Context/Use case:
Detecting two objects in the same frame using FOMO model using iPhone camera during live classification

Summary:
When two different objects were shown in the frame at the same time, the model wasn’t able to detect anything

Expected Results:
Originally, I expected for both objects to be detected in the frame

Actual Results:
When both objects were shown together, none of them were able to be detected

Reproducibility:

  • Always

Environment:

  • Platform: [iPhone camera]
#2

Hi @rakshana

You should be able to track two different classes with FOMO without issue, this looks like a poor dataset / class imbalance issue. Try to follow our model improvement guide: Increasing model performance | Edge Impulse Documentation

Best

Eoin

#3

Thanks, Eoin for the suggestions! I will try these methods in my project.

#4

Hello @rakshana,

I just had a quick look at your project and my suggestion would be to:

  • add more images to your dataset
  • also add images with both the banana and the cup in the same frame

Best,

Louis