Nicla vision object detection false detection

My Model is 93.8% F1 score but when deploy to my Nicla Vision board. I have false detection of my model all the time. What is the problem in my model. I haved watched the video that explain FOMO https://www.youtube.com/watch?v=VzJZM5p24Tc&t=799s

My project:

What’s wrong with my datas

Best regards

Martin

Hi @mlasalle

Nothing is wrong with your data, you just dont have anywhere near enough :blush:

You only have 78 items and the dataset is imbalanced, if you have a 93.8% accuracy that is only for the very small set of images that you trained on. Add more MUCH more to expect better results, dont be surprised if the % of accuracy goes down, that just means you need to be more consistant with the capturing of images, distance, background, and other factors. To get a model that can generalise better. :chart_with_upwards_trend:

Please see our getting started guide for some tips:

Best

Eoin