This is a naive approach to demonstrate how to use a multiclass classification to identify the breed of a cat.
It is NOT a realiable solution and it has no intention to be one.
The motivation of this work is at here:
Before you run
Please update the 2 TSV files:
- cats.tsv: under folder
- test.tsv: under folder
You have to specify the absolute path to match your local machine where you store those images.
The trained model is over 120MB so I cannot push to git. So you have to train the model locally first.
There are 2 ways to run the console app.
1. Use VS.NET community (easier)
Debug the CatTrackerML.ConsoleApp with an argument
train, which will use some default arguments.
Next debug the CatTrackerML.ConsoleApp with an argument
predict, which will again use some default arguments.
2. Build it with command line
Navigate to the root folder, and run this:
dotnet publish -c Release -r win10-x64
This will create the exe file under a folder like:
You can copy all files from the subfolder
netcoreapp2.1, so you can run it with default arguments; otherwise try
CatTrackerML.ConsoleApp.exe -h CatTrackerML.ConsoleApp.exe train -h CatTrackerML.ConsoleApp.exe predict -h
Each will explain how to use. It's mostly vanila ML.NET stuff.
All the images are randomly downloaded from internet. If any violate any copyright, please let me know and I am happy to remove them.