Urban Waste Dataset for Pakistan

This dataset contains 3,693 images of garbage and non-garbage scenes collected from 8 cities in Pakistan (Islamabad, Rawalpindi, Karachi, Lahore, Sialkot, Quetta, Multan, and Mirpur AJK). The dataset is divided into Train (70%), Validation (20%), and Test (10%) splits, with two categories: Garbage and Not Garbage. To our knowledge, this is the first curated and open-access garbage image dataset from Pakistan, created to support research in computer vision, deep learning, and smart city waste management. Images were collected manually, supplemented with targeted web searches and social media scraping. All data was reviewed for quality, and sensitive details (faces, license plates) were blurred. A Metadata.csv file provides additional annotations, including filename, label, split, width, height, file size and file format. Intended Use: Image classification, object detection, waste detection, machine learning, deep learning, Pakistan urban waste research, smart city applications, and environmental monitoring.License: Released under [CC BY 4.0].Citation: Please cite as Mahmood, I. (2025). Garbage Detection Dataset for Pakistan. Zenodo. DOI: 10.5281/zenodo.17025661.