FOMO is an object detection model developed by Edge Impulse which is based on MobileNetV2. However, this repo will provide Squeezenet, MobilenetV3 and MobileViT variants as well as MobileNetV2.

git clone git@github.com:bhoke/FOMO.git

For Tensorflow implementation, your custom dataloader should consume the dataset accordingly.

or you should write your own data loader class which can be eithe keras dataloader or tf.data.Dataset class.

When data loader is ready, you can train your model using: python train.py

Finally, you can run inference on any image using:

python predict.py

• MobileNetV2
• MobilenetV3
• MobileViT

Precision and hardware-specific metrics will be added here soon.