Fingervein Recognition Model3

---

This project implement in matlab and python,which be finshed in 2017.5.Be sure you have a GPU first,preprocess and cnn model in cpu is slow.

01_Genetic_Preprocess

This part use Genetic algorithm to finish remapping. Roi detect and clip should be done first.(important). You can use your own database to generate preprocess img. Tip:Genetic algorithm is expensive and your can use GPU to do the job or cut down the number of iterations

02-Feature_Extraction

Sparse auto encoding is doing in this part.

03_Convolution&Pooling

A simple CNN model is used in this part.After this part,get the feature map(Traub Features). You can use it in 04 or a new tensorflow model.

04_Classification&Test

Use OneVsAll classofocaion method.

Matlab

IDE:Matlab 2016a The minFunc lib is need for this project, it can be found in Stanford course.

Run

Please add this project into PATH and run Main.m;

use your own dataset

In model 01_Genetic_Preprocess; Change the path in Main_1.m & Main_2.m and Main_LoadData,m;

Tensorflow

python2.7 + tensorflow1.1

Model

Python is simple,and tensorflow is very convenience.After model 03,you can use different model in tf to improve it.A simple example will be include in my antoher warehouse.

The dataset is very large.

Data-Set

• 1. The first Data-Set can be find hereSDUMLA_HMT, Release Aggrement should be offer.
• 2．The second Data-set can be find here fvc-2004;You can also find it in other place. It's a open source database. You can generate feature from code.