PCA-SVM-Features-based-Face-Images-Classification

KO | EN

gitlite — search

#javascript #python #golang #go #android #machine-learning #macos #java #awesome #cli #awesome-list #pytorch #security #mcp

PCA-SVM-Features-based-Face-Images-Classification

(★ 8)

Face recognition on The Database of Faces at a glance. Data pre-processing. Dimension reduction using PCA. Support Vector Machines (For classification)

PCA-SVM-Features-based-Face-Images-Classification 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# PCA-SVM-Features-based-Face-Images-Classification

### Objectives:

Face recognition on The Database of Faces at a glance.
    Data pre-processing.
    Dimension reduction using PCA.
    Support Vector Machines (For classification)

### Project work flow:

![Work Flow](https://raw.githubusercontent.com/afaq-ahmad/PCA-SVM-Features-based-Face-Images-Classification/master/pics_for_presentation/Work_Flow.JPG)

### Data-Set(The Database of Faces at a Glance):

The data-set consists of 400 pictures of 40 people.
    Each picture has a dimension of 112X92 pixels.
    Data set comprises of pgm format.

Front faced pictures of a single subject was captured in different perspective & can be seen by plotting few examples of the given data-set.

![sample](https://raw.githubusercontent.com/afaq-ahmad/PCA-SVM-Features-based-Face-Images-Classification/master/pics_for_presentation/sample.jpg)

### Data pre-processing:

Each image is flattened into row vector of size 10304.
    The size of whole data set becomes 400X10304.
    Then converted into a numpy ndarray.

### Dimension reduction using PCA:

Redundant features are removed using PCA.
    Eigen vectors with more variance are kept while with low variance are discarded. 
    Output matrix from PCA will look noting like original image.

### Support Vector Machine(SVM):

First of all for classification SVM is applied directly without using SVM.
    To get maximum margin, Linear classifier was used.
    100% accuracy was achieved.

### Score without using PCA:

When original numbers are used that is dimension reduction isn’t performed we have maximum score.

![score without pca](https://raw.githubusercontent.com/afaq-ahmad/PCA-SVM-Features-based-Face-Images-Classification/master/pics_for_presentation/score_Without_pca.jpg)

### SVM after using PCA:

>PCA was implemented for K=1,2……10
    >For K=1 , error was 94%.
    >For K=10, error was 1%.
    >Significant improvements made when dimensions were increased.

### Results with PCA:
As we are increasing number of dimensions error is decreasing and score is increasing to one.

![Result with Pca](https://raw.githubusercontent.com/afaq-ahmad/PCA-SVM-Features-based-Face-Images-Classification/master/pics_for_presentation/result_With_pca.jpg)

PCA-SVM-Features-based-Face-Images-Classification

PCA-SVM-Features-based-Face-Images-Classification 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

PCA-SVM-Features-based-Face-Images-Classification

PCA-SVM-Features-based-Face-Images-Classification 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README