maskedClassification
A coreML test for machine learning focusing on distinguishing masked faces or not.
1. DataSource for Machine Learning.
ffmpeg -r 1 -i http://videoName.mov -r 1 "$filename%03d.png"
Record a video maybe 10s or 15s, then extract each key-frames as a picture at every second in ffmpeg
Take the pictures as the datasource for Machine Learning.
2. Trainning, Validating and Testing.
With the help in Xcode, we can create a client-side lite-coreML if we use [CreateML] in [File]--[Open Developer Tool]
3. Export the MLModel for integrating in existed projects CoreML.framework etc. That's awesome but obviously OVERFITTING.
4. Go alittle deeper for ML for adjusting.(To be continue.)
