Comparison of image classification web services: demonstration and implementation

OData support
Supervisor:
Dr. Pataki Béla József
Department of Measurement and Information Systems

Nowadays applications using artificial intelligence are common. It is not only used in software which is specifically designed for intelligent behaviour but also in all kinds of mobile and web applications. One of the most popular use cases of the technology is image classification. Image classification means, that the application is capable of identifying an image’s content without human interference.

By searching the internet, it is not too difficult to find numerous image classification services, which are able to execute the discussed task with varying precision. However, it is far from trivial to decide, which of the services is the most precise and the most reliable.

It is possible to measure this reliability with the test application implemented for the thesis by using cutting edge web technology. The web application provides a tool for the user to upload images and evaluate the responses given by the various image classification services. The goal of the thesis is to describe the technologies used in the development of the test application, discuss the potential use cases of AI-based image classification and measure the precision of the used services.

After evaluating the results, an order of the services can be set up in the aspect of reliability and precision by calculating the number of true positive and false positive responses.

In the future, the application can be potentially used not only for testing purposes but also for production use, however, to achieve this, expanding its features is crucial. By doing so, the result is an intelligent gallery app, which can automatically categorise the uploaded images and provide the user with a search tool where they can find images by content.

Downloads

Please sign in to download the files of this thesis.