Defect Detection is a method where automated visual evaluation systems can be used to control the product quality in an industry to meet its objective.
During various stages of a manufacturing process or product lifecycle, surface level defects may occur, such as missing and misaligned components, physical damages like scratches, cracks, holes, etc. To find these defects, either visual inspection is done by human experts, or Automatic Optical Inspection (AOI) is conducted using computer algorithms. Surface analysis technique is an algorithmic attempt to inspect a surface for possible defects. Defects could be of various types and many different methods are used for inspection.
This work is a demonstration of a potential wood industrial project involving the development of AOI techniques in a manufacturing or recycling environment. The set-up potentially could contain a high quality camera which sends images of components to a computer running defect detection algorithm.
Problem Statement: Anomaly Detection and Localization
MVTec Anomaly Detection (MVTec AD).
Fine-tuning training for the wood category of this dataset was done using Google Colab.
This repository utilizes the following model:
PaDiM stands for Patch Distribution
Modeling and it generates patch embeddings for anomaly localization.
It makes use of a pretrained convolutional neural network (CNN: ResNet-18) for embedding extraction
and has the two following properties:
- Each patch position is described by a multivariate Gaussian distribution
- PaDiM takes into account the correlations between different semantic levels of a pretrained CNN
The following features are embedded within the model and anomalib package:
- Embedding Extraction
- Normality Learning
- Anomaly Map Computation
- Anomaly Score using Mahalanobis distance where the final anomaly score of the entire image is the maximum of the anomaly map M
- git clone/ download zip
- cd into dir
pip install -r /path/to/requirements.txt
- Dashboard
python main.py
Repository is containerized.