/Anomaly-Detection

Anomaly detection, Convolutional VAE, Convolutional AutoEncoder

Primary LanguagePython

Anomaly Detection

Introduction

This code originally was written for turbine diagnosis.
For diagnosis, our team used feature image that extracted from blade vibration signal was used as training and test data.
But, unfortunately the detail explanation and dataset release are trade secret. therefore, this code is operated by replacing the dataset with open dataset.

The code related to turbine diagnosis is preserved on BVMS branch, master branch is related to open dataset.

Open dataset

It is an image dataset provided by AI challenge website called DACON. First, this is consisted of 6 types of crop images.

  • Strawberry, Tomato, Paprika, Cucumber, Pepper, Grapes.

Second, 12 types of crop status and 4 types of degree of disease damage are consisted.

Requirements

The code on BVMS branch was tested with Tensorflow 1.12.0, CUDA 8.0 and Ubuntu 18.04

Cleaning up the code