This is the project I did for training a custom image classifier to detect the presence of a COVID mask as a cool Machine Learning problem to solve. My goal was to run it on a small device that could operate at a very low cost as an egress point to make sure for examples that doctors and nurses were compliant to protect themselves entering a room that could contain a hazard and masks were required.
I started running it locally on an older graphics card but getting the libraries set right was a chore
Full Google Colab for Training Model with Tensorflow v1.13 and convert to tflite and EdgeTPU https://colab.research.google.com/drive/1uEkP5j7KM9eSkCUtUyauy7-Dyd5XAY6e?usp=sharing
Google Drive with the Images folder and Config for the Google Colab
https://drive.google.com/drive/folders/1vmWE60hk2m1-LpgTiaFPef6ypZKsBlUi
Original video with walkthrough of training on Windows 10:
https://www.youtube.com/watch?v=Rgpfk6eYxJA
Github repo for Training
https://github.com/EdjeElectronics/TensorFlow-Object-Detection-API-Tutorial-Train-Multiple-Objects-Windows-10
My repo for downloading and resizing images
https://github.com/contractorwolf/googleimagedownloader
Image Labeling software
https://www.dropbox.com/s/tq7zfrcwl44vxan/windows_v1.6.0.zip?dl=1
https://github.com/tzutalin/labelImg
How to run Tensorflow Lite on Raspberry Pi
https://www.youtube.com/watch?v=aimSGOAUI8Y
Using the Google Coral TPU with Tensorflow Lite
https://www.youtube.com/watch?v=qJMwNHQNOVU