ML.NET project to detect sliced from unsliced food via an image classifier.

For building a model:

dotnet SliceClassifier.dll build

For making predictions with a trained model:

dotnet SliceClassifier.dll predict -m path/to/model/MLModel.zip

For help about parameters:

dotnet SliceClassifier.dll build help

dotnet SliceClassifier.dll predict help

• Model training options must be changed inside the code, on ModelBuilder.cs.
• When no parameters are used the dafaults values in the code are written to make everything work. Except for the model ZIP that needs to be placed next to the DLL.
• To this day, I could not debug the project in Visual Studio, had to run the generated DLL to read the errors.
• Best obtained model is MLModel.zip, with labels named Sliced and NonSliced (instead of Unsliced). However, new models will be generated next to the executable.
• Python script rename-pics.py is just an auxiliary script to change the name of the pictures in each data set folder when adding new ones.