Glass vs No-Glass Classification

Pattern Recognition & Machine Learning (CSL2050) Major Course Project Developed under the guidance of Dr. Richa Singh

Data

• Google Drive link
• Data from Kaggle competition

Loading data?

Import Data from Kaggle. It has train features, train label and test labels. For MLP , KNN , SVM we can directly use these features (latent vectors). For CNN we produce images by Latent vector.

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 (imoprt neccesary libraries and data)
• Data Preparation (Latent Vector to image. Separating these images into different directories with respect to their label and train/val/test.)
• Data Preprocessing (For images make training testing set by ImageDataGenerator. For latent vector split into train/test , standardized data and reduced dimentioned data.)
• MLP , KNN , SVM (Three Classification model. Each model has 3 set {1. Noraml , 2. Dimensionality Reduction , 3. Feature selection})
• CNN (A three layer CNN model for image classification)

Demonstration

The video refrence to code.(Drive link)