Comparative analysis of deep and transfer learning techniques for two and multi-class weather image classification

Authors

Saumya

Issue Date

2021

Degree

MSc in Data Analytics

Publisher

Dublin Business School

Rights

Items in eSource are protected by copyright. Previously published items are made available in accordance with the copyright policy of the publisher/copyright holder.

Abstract

Information about the weather plays a very vital role in day-to-day life of humans, including various sectors like agriculture, business, traffic etc., knowing the weather beforehand helps to resolve several weather-related issues, this generates need of implementing a robust model to find weather type by analysing weather images. Very few research has been conducted in this field using Transfer Learning. This research fills the gap by using pre-trained models and CNN model. The focus of this research is classification of different types of weather using transfer learning. Image2Weather dataset taken and divided into two datasets, Weather4Class having four weather classes and Weather2Class dataset having two weather class and four deep transfer learning models applied further performance is compared. The Data Augmentation and Feature Extraction technique is applied on VGG16. InceptionV3, DenseNet and CNN were also considered for accuracy comparison. The research concluded VGG16 model performed best compared to other techniques.