A.G.E ( Age-Gender-Ethnicity ) Prediction

This project has been done as part of the course CS337 (Artifical Intelligence and Machine Learning)
The aim of this project is to build the best possible models for Age, Gender and Ethnicity prediction using several Machine Learning techniques

Important Techniques Used

• Face Detection
• Convolutional Neural Networks
• Boosting
• Bagging

How to run code ?

Training Process

First create the pickle files for image dataset using create_pickle.py

Then use the file train.py for training the models, the usage has been shown below

usage: train.py [-h] [--model MODEL] [--lr LR] [--dropout DROPOUT] [--batch_size BATCH_SIZE] [--epochs EPOCHS] [--n_estimators N_ESTIMATORS] [--backbone_output BACKBONE_OUTPUT]

You can also refer to Scripts/data_gen.sh file for training all models simultaneously

NOTE : For training of Bagging part, we had to train several single model because of the memory limit within the gpu and some cuda errors. So, while evalating it we have combined all separately trained models.
In order to train the Bagging related models you can look up the bagging_train.sh bash script

Testing Process for Test Dataset

Use test.py for testing the models on whole image test dataset, it will also generate the corresponding plots for all the models

NOTE : You might want to modify the model's list withing test.py file if you want to test and plot graphs for some specific data

Demo

Use run.py for predicting the age gender and ethnicity for a given sample image

python3 run.py <sample-image>