This repository contains a CNN implementation for quality detection of ring images.
This repository contains two main scripts: train.py
and test.py
.
This script serves to train our model, log the process' metrics and export a checkpoint. Eventually a test on unseen data is performed for evaluation.
A script for evaluation an existing model on a dataset. It is possible to test the performance of the model.
To avoid leakage from training data, please indicate a new unseen data in the following way:
$ python test.py --data_path <your new data path>
For loading the model (.pb file):
$ python test.py --model_path <trained model .pb file>
The training process shown good performance as in the following image: