GenderClassification

The aim of this project is to construct a Convolutional Neural Network (CNN) model that is capable of accurately distinguishing between male and female categories, given an image of a human.

Project Structure

├───App
│   ├───static
│   │   └───faces
│   ├───templates
├───Code
├───Model
└───Pic

Methodology

To prepare the images for classification, OpenCV (CV2) is utilized to crop faces from the images. The classification process is based on a simple CNN algorithm

Data

Dataset on kaggle

• Training :47009

• Validation: 11649

Note : This set is splited to : 2,336 for testing and 9,313 for validation

• Samples:

Training

Augmentation

• RandomFlip("horizontal")
• RandomRotation(0.1)
• RandomZoom(0.1)
• RandomBrightness(factor=0.2)
• RandomContrast(factor=0.2)

Callbacks

• LearningRateScheduler
• EarlyStopping (monitor='val_loss', patience=3)

Model Architecture

Compilation Parameters

• Adam : initial_lr = 0.0001, Drop = 0.1, Every 10 epochs
• loss : Categorical Cross Entropy
• metric : 'Accuracy'
• epochs : 19

Training and Validation Matrices

• Tranining loss : 0.0904
• Training Acc : 0.9680
• Validation loss : 0.0802
• Validation lAcc : 0.9695

results

• Accuracy: 1.0
• Precision: 1.0
• Recall: 1.0
• F1-score: 1.0
• Confusion matrix:

Links:

Model File

Simple Flask Web App

Demo