Age Group and Gender estimation from face image using CNN.

Two Custom CNN layers are trained for age group and gender estimation.

For the Age Group classification trained CNN over the 23000 images, and for the Gender Classification trained CNN 4000+ images of indian face.

Age Group parameters are :

Kid = 0 - 14,
Youth = 15-40,
Middle Age = 41-60,
Senior = 60+.

Gender parameters are :

1) Male
2) Female

The name of dataset is "UTKFace" where the information about age, gender, and ethnicity is given in the image title. Link: https://www.kaggle.com/jangedoo/utkface-new

Usage

1. Install Following libraries using pip

• numpy 1.16.2
• opencv-python 4.0.1.24
• face-recognition 1.2.3
• tensorflow 1.11.0
• tensorflow-gpu 1.11.0 (Optional if you have Nvidia CUDA Supported GPU)
• keras 2.2.4
• PyQt5 5.12.1

2. Clone or download the repository and extract it into a folder and open it.
3. Run gui.py using cmd

python gui.py

Convolution Neural Network

Link: https://medium.com/@RaghavPrabhu/understanding-of-convolutional-neural-network-cnn-deep-learning-99760835f148

Training CNN

Image Preprocessing:

Face Images are resize into 180X180 pixel size and converted into gray scale and given image input to CNN.

For more than 2 class classification label is encoded with one hot encoding.

Results

Confusion matrix for the Gender classification:

Confusion matrix for the Age Group classification: