An Image Dataset for Illegal Dumping Detection in Urban Areas of Bangladesh

Description

This dataset is developed for detecting illegal dumping using machine learning in urban environments. The dataset contains two main classes: Positive and Negative. The Positive class represents images containing visible dumped waste or illegal dumping, while the Negative class represents clean urban scenes without dumping. The original dataset contains 804 real images, including 326 Positive/Dump images and 288 Negative/Clean images and 200 Critical Clean images. The real images were divided into training, validation, and testing sets using a 70/20/10 split. To improve model generalization, additional augmented images were generated only for the training set. Validation and test sets contain only real, non-augmented images to ensure fair model evaluation. The final prepared dataset contains 3,557 images. Please cite our research paper (link is provided below) if you use the dataset.

Steps to reproduce

Please cite our research paper (link is provided below) if you use the dataset.

Institutions

• University of Dhaka

  Dhaka Division, Dhaka

Categories

Licence