# Geospatial dataset on deforestation and urban sprawl in Dhaka, Bangladesh: A resource for environmental analysis

## TL;DR

This paper introduces a high-resolution dataset tracking deforestation and urban growth in Dhaka, Bangladesh, to support environmental research and machine learning applications.

## Contribution

The paper presents a novel, annotated geospatial dataset for studying deforestation and urban sprawl in Dhaka using machine learning techniques.

## Key findings

- The dataset includes annotated satellite images and JSON masks for tracking tree cover changes over a decade.
- It supports machine learning tasks like object detection and change detection for environmental analysis.
- The dataset is useful for education and research in environmental science and image processing.

## Abstract

This dataset comprises high-resolution satellite images for monitoring deforestation in Dhaka, Bangladesh. Data were acquired via Google Earth Pro from fixed locations to maintain consistency in observing tree cover alterations. Each image includes annotations and JSON mask files that delineate tree cover and deforested areas. The dataset facilitates machine learning applications, including object detection, semantic segmentation, and change detection. Image resolution and aspect ratio vary, with 5-35 images recorded per location annually over a decade. This data serves researchers investigating urbanization's environmental impact and the gradual reduction of tree cover in a rapidly evolving urban environment. Utilizing the annotations and masks enables the training of machine learning models to identify and forecast vegetation changes, aiding environmental monitoring and conservation initiatives. Furthermore, the dataset is readily applicable for educational purposes in disciplines such as geography, environmental science, and machine learning. It provides critical insights into the application of machine learning and image processing in addressing real-world environmental issues.