VARUN THAZHATHARACKAL

Comparison of different methods for feature extraction and selection for hyperspectral images (HSI) of different types of demolition waste for efficient automatic sorting, - Different methods of dimensionality reduction techniques for HSI to generate a feature vector, - Feature Vector as a input for Machine Learning methods to classify the different material types in demolition waste, - Development of the model using MLFlow, - Investigating how many, and which features have the highest recognition rate.

- Hyperspectral image acquisition for the dataset and adding files in Blob storage triggers an event in Azure Event Grid - Azure functions calls Azure Computer Vision(CV) API to analyze the newly uploaded image - CV accesses the image via the blob URL parsed by Azure functions - Azure functions persists the CV API response in Azure Cosmos DB. - Response results of analysis, along with the image metadata - Retrieves the results of classification

Machine Learning, Phogrammatic Computer Vision, Image Analysis, Algorithms and Data Structures, Statistics, Optimization