search

Jaluus / 2DMatGMM

Code release for "An open-source robust machine learning platform for real-time detection and classification of 2D material flakes"
https://flakedemo.uslu.tech
MIT License
33 stars 8 forks source link

Needs retraining or not? #1

Closed cl-love-mtt closed 3 months ago

cl-love-mtt commented 3 months ago

I am also trying to find thin layers of graphene, if the light path or silicon conditions are different and I get a picture with a yellowish color, do I need to retrain the model?

Jaluus commented 3 months ago

Hey!

Yes, If your silicon has a different thickness, i.e. not 90nm, your camera has a different pixel sensitivity curve or you have a lamp with a different spectral profile, I would retrain the model with a few images. About 3 to 5 per tickness should suffice.

If you are interested in optimizing the contrast of your graphene flakes, a good paper to read would be "Making Graphene Visible".

They test for 300nm SiO2 but our lab uses 90nm due to stronger and more stable contrast in the Red Green and Blue channel of the Camera.

If that works for you, I would love to hear how it went!

Best Jan