Skip to content

Latest commit

 

History

History
18 lines (11 loc) · 716 Bytes

README.md

File metadata and controls

18 lines (11 loc) · 716 Bytes

Defect classificaiton using NEU dataset

A simple, binary classification using Xception model which was pretrained and of which the weights file can be fetched from Keras

Dataset

NEU surface defect database

Samples with surface defects are merged into a single category (folder) call MT_Defects.

Cleaned-up dataset is stored in folder data

Data preparation

Call: python data_prep.py after modifying corresponding folders and split percentages for train/val/test.

Train/test

Call: python defect_classify.py

For parameters see inside the code.