![](https://csdnimg.cn/release/download_crawler_static/89017620/bg1.jpg)
INSTALLATION / USER MANUAL
AUTHOR: SUNDY
DEYE: KEEP AN EYE ON DEFECTS INSPECTION
30/10/2020
COPYRIGHT@2020-2025
![](https://csdnimg.cn/release/download_crawler_static/89017620/bg2.jpg)
Abstract
Defect Eye (DEye) is a deep learning-based software for manufacturing sur-
face defect inspection. It provides the basic function modules to facilitate
the development of different defect inspection applications. The applications
cover the full rang of manufacturing environment, including incoming pro-
cess tool qualification, wafer qualification, glass surface qualification, reticle
qualification, research and development. Also, It can be used for medical
image inpsection, including Lung PET/CT,breast MRI, CT Colongraphy, Dig-
ital Chest X-ray images. This software library contains the basic function
modules about data processing, model training and model inference. It is de-
veloped to reduce the burden of programmers who works in this field. Based
on this software, developers can design the added functions according to their
requirements. This document describes how to build DEye software and how
to use it.
1 Introduction
The DEye [
4
] software library consists of 7 modules with some inter-dependence. The function
of each module is shown as below:
• GUI
: a simple user interface for model training and inference (programmed with C#).
• trainer
: it contains data processing and model training based on Tensorflow frame-
work (programmed with C++).
• classifier
: it contains model inference when feed the input images (programmed with
C++).
• dataGen
: a module to generate training data under the specified size (programmed
with C++).
• example: some function test cases (programmed with C++).
• SVM
: a simple demo to show how to use SVM classifier to classify images (programmed
with C++).
• txt2xml: a tool to convert txt format to xml format (programmed with C++).
2 Development Environment
The Microsoft Visual Studio 2017 (VS2017) is used to develop the DEye software under Window
(English version, 64 bits) operation system, the version of VS2017 is show in Figure 2.
Besides, the configuration of each module is shown in Figure 3 and Figure 4. All of these
modules should be configured in a right way.
3 3rd Party Library
DEye is depended on several third party libraries, i.e., Tensorflow [
1
], OpenCV [
2
], Glog [
3
],
which all of them are open source. In this project, the C++ version of Tensorflow 1.4, OpenCV
3.4, and Glog are used to support DEye library development. The newest 3rd-party libraries
should be built by developers who want to update the version. Fortunately, the default version
is provided with the following links:
• BaiduYun Link: https://pan.baidu.com/s/1o9tv1n8, password: ekec
•
Google Link: https://drive.google.com/open?id=1kANDNErMNLU9wNR3rKhUTz–
ltwPNPUv
All the libraries should be downloaded into the computer, then configuring the environment
with the following steps:
1
![](https://csdnimg.cn/release/download_crawler_static/89017620/bg3.jpg)
Figure 1: Visual Studio 2017 version
Figure 2: Configuration of GUI module.
2