Is the FOMO a pretrained model? How can we use FOMO with MobileNetV2?

Help
#1

Question/Issue:
Is the FOMO a pretrained model? How can we use FOMO with MobileNetV2?

#2

Hi @soujanyanaidu25,

FOMO starts as a pre-trained model but is adjusted as needed to generalize to your training data (i.e. it uses “transfer learning”). FOMO is made with MobileNetV2, but it is modified to perform basic object detection on a grid. I recommend checking out my video here to learn more about how FOMO works: tinyML Talks: Constrained Object Detection on Microcontrollers with FOMO - YouTube

#3

Also there are some more details about the architecture available in the docs FOMO: Object detection for constrained devices - Edge Impulse Documentation

FOMO uses a MobileNetV2 backbone, but could use anything you like really.

1 Like