The aim of this project is to use deep learning to classify 6 possible defects on metal surfaces:
- 0: Crazing
- 1: Inclusions
- 2: Patches
- 3: Pitted Surface
- 4: Rolled in Scale
- 5: Scratches
Two types of datasets were used in the project: the first with only 90 images (15 for each defect: 60 in the training set and 30 in the test set) and the second, with which the best results were achieved, with 1800 images (300 for each defect: 1620 in the training set and 180 in the test set).
The whole project was carried out entirely in Python can be found on the Jupyter Notebook with Python code, output and comments.