转自：https://bbs.cvmart.net/articles/571

DEye (Keep an Eye on Defects Inspection)

1. Abstract

Defect Eye is an open source software library based on tensorflow1.4, which focus on surface defect inspection. The application area cover the full range of yield applications within the manufacturing environment, including incoming process tool qualification, wafer qualification, glass surface qualification, reticle qualification, research and development, and tool, process and line monitoring. Patterned and unpatterned wafer defect inspection and qualification tools find particles and pattern defects on the front surface, back surface and edge of the wafer, allowing engineers to detect and monitor critical yield excursions. Also, It can be used for medical image inpsection, including Lung PET/CT,breast MRI, CT Colongraphy, Digital Chest X-ray images.

 

2. Usage

Compiled tensorflow-r1.4 GPU version using CMake,VisualStudio 2015, CUDA8.0, cudnn6.0.

• tensorflow.dll, tensorflow.lib, libprotobuf.lib
• Baidu Cloud Link：https://pan.baidu.com/s/1o9tv1n8 password：ekec
• Google Drive Link: https://drive.google.com/open?id=1kANDNErMNLU9wNR3rKhUTz--ltwPNPUv

How to use DEye

• Install VisualStudio Community2015 Install NVIDIA CUDA 8.0
• git clone https://github.com/sundyCoder/DEye
• Download tensorflow.dll, place it under DEye/bin
• Download tensorflow.lib and libprotobuf.lib, place theme under DEye/extra/tensorflow-r1.4/
• Download inception_v3_2016_08_28_frozen.pb, place it under DEye/data
• Open Visual Studio Solution "DEye.sln" which should be under DEye/build/vc14，Solution configurations option choose "Release", Soluton Platform option choose "x64".
• Build and run the GUI project, you can do model training for your inspection cases.

3. Applications

3.1 IC Chips Defects Inspection

3.2 Highway Road Crack Damage Inpection

3.3 Fabric Defects Inpection

3.4 Cover Glass Inpection

3.5 Civil Infrastructure Defect Detection

3.6 Power lines Crack Detection

3.7 Medical Image Classification

4. Datasets

1. Weakly Supervised Learning for Industrial Optical Inspection

   DAGM: https://hci.iwr.uni-heidelberg.de/node/3616

2. Micro surface defect database

   https://pan.baidu.com/s/1QM0AxlGjUlkHHyxwamIMmA

3. Oil pollution defect database

   https://pan.baidu.com/s/1_aU_Bfh7lcxpYW1no2MlUQ

4. Bridge Crack Image Data

   https://pan.baidu.com/s/1bplPrPl

5. ETHZ Datasets

   ETHZ: https://www.vision.ee.ethz.ch/en/datasets/

6. RSDDs dataset

   https://icn.bjtu.edu.cn/Visint/resources/RSDDs.aspx

7. Crack Forest Datasets

   https://github.com/cuilimeng/CrackForest

8. CV Datasets on the Web

   https://www.cvpapers.com/datasets.html

9. An RGB-D dataset and evaluation methodology for detection and 6D pose estimation of texture-less objects

   https://cmp.felk.cvut.cz/t-less/

10. CV Dataset on the web

    https://cvpapers.com/datasets.html

5. Contact

• Email: sundycoder@gmail.com
• QQ: 1316501606

Notice: Any comments and suggetions are welcomed, kindly please introduce yourself(name, country, organization etc.) when contact with me, thanks for your cooperation.

6. TODO List

• A user-friendly GUI ( welcome to contact with me if you want to be a collaborator)

7. License

Apache License 2.0

8. Citation

Use this bibtex to cite this repository:
@misc{DEye,
title={CNN-based Manufacturing Defect Detection on Tensorflow},
author={Sundy},
year={2017},
publisher={Github},
journal={GitHub repository},
howpublished={\url{https://github.com/sundyCoder/DEye}},
}