Mask-vs-NoMask-Classification

Minor Course Project, Pattern Recognition & Machine Learning (CSL2050), developed under the guidance of Dr. Richa Singh

Data

• Google Drive link

Loading data?

Import Data from Resource given. It has train features, train label and test labels. For RF, MLP , KNN , SVM we can directly use these features (latent vectors) generated by extracting features from images. For CNN we use images available in dataset.

How to Execute?

The collab notebook is written sequentially. Just run the blocks one by one. Some block may even take few hours to run.

Structure of Code

The code is divided into multiple selection.

• Dependencies & Data Loading (import neccesary libraries and data).
• Latent Vector creation (Image to Latent Vector by feature extraction).
• Data Preprocessing (For images, make training testing set by ImageDataGenerator. For latent vector split into train/test, standardized data and reduced dimensioned data).
• RF, MLP, KNN, SVM (Three Classification models. Each model has 3 set {1. Normal Data, 2. Standardized Data , 3. Feature selection}).
• CNN (A three layer CNN model for image classification).

Demonstration

You can check the project report here