UGV-NBWASTE: An oriented dataset for non-biodegradable waste in BangladeshMendeley Data
The “UGV-NBWASTE” dataset is built for those who manage non-biodegradable waste. The selection of non-biodegradable waste has been decided on adverse environmental conditions, particularly waste management in landfills and water. The dataset is collected from the Barisal district of Bangladesh, wher...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-06-01
|
| Series: | Data in Brief |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2352340925002914 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The “UGV-NBWASTE” dataset is built for those who manage non-biodegradable waste. The selection of non-biodegradable waste has been decided on adverse environmental conditions, particularly waste management in landfills and water. The dataset is collected from the Barisal district of Bangladesh, where eight distinct types of waste (Plastic Bottle, Hard Plastic, Mask, Medicine Packet, Packet, Polythene, Cocksheet, and Plastic Sandal) are selected based on their widespread use, durability, and difficulty in recycling or managing them via conventional waste disposal methods. Furthermore, waste images are captured using smartphones in indoor and outdoor situations, such as floating in water or partially buried in the mud, which is crucial to diversifying the dataset for effective detection and classification. After data collection, various techniques are applied during the image pre-processing stage to significantly improve the quality of the original images. These include Image Quality Assurance (i.e., image verification and image cleaning) and Image Enhancement (i.e., brightness normalization and image resizing). Then, all images are annotated in oriented bounding box (OBB) format, ensuring waste detection at different angles. The total number of original images is 3600. Waste can be reliably identified whether it is flat, crumpled, or partially obscured, which guarantees the dataset's ability to identify waste in different circumstances and orientations. |
|---|---|
| ISSN: | 2352-3409 |