An AI-Powered Web Application for Waterfall Recognition and Eco-Tourism Enhancement in Sri Lanka: Falls Explorer

Abstract

This research presents the development of Falls Explorer Sri Lanka, a mobile-responsive web application that uses artificial intelligence for automatic waterfall recognition. The core innovation lies in applying a custom-developed convolutional neural network (CNN) to classify waterfall images based on their visual features. A custom image dataset was created by collecting and organizing photos of popular waterfalls in Sri Lanka, and the model was trained using TensorFlow. The custom CNN model achieved 92% validation accuracy after 25 epochs of training, with inference times under 1 second per prediction. The system successfully classified waterfall images across 20 different waterfall classes with precision scores ranging from 88% to 95%. Users upload a photo of a waterfall through the interface, and the system returns the predicted waterfall name along with travel details from a local JSON database. In addition to the recognition feature, the platform offers comprehensive functionalities such as displaying detailed waterfall information (name, location, description), listing nearby hotels, showing current weather forecasts for safe travel planning, hosting a community forum for users to share experiences and images, providing a carbon footprint calculator to estimate travel impact, and an interactive location search map to explore specific sites manually. This solution bridges the gap between technology and ecotourism, supporting conservation-friendly tourism by enabling travellers to appreciate natural attractions without invasive markers or infrastructure.