Simple Color Classifier with TensorFlow (Python).
The classifier is not so accurate (80% accuracy, depending on the size of the dataset and the noise) because no filter was applied since this is just a simple example. You can add your own filter. Remember this is a classifier based on human perception, not on a mathematical calculus, that's why noise is not easy to remove.
This project was inspired by this youtube playlist.
Please consider supporting the author of that playlist.
The project is based on a crowdsourced database and it uses TensorFlow to build the model and process the data. I also used Tkinter to build a (VERY) simple (quite bad) GUI, just to test the model with user-given data.
More on Tensorflow: https://www.tensorflow.org/
More on Tkinter: https://wiki.python.org/moin/TkInter
If you don't need the GUI, this is the right order to execute the scripts in order to get the data, process them and create and save the model:
-
getData.py
-
processData.py
-
train.py