Tobacco and Weed Detection Dataset

TobaccoWeed Dataset: Object Detection for Agricultural Field Images

This dataset contains 307 high-resolution images of tobacco fields with manually annotated bounding boxes for tobacco plants and weeds. The images were captured in Erbaa district, Tokat province, Turkey, in August 2023 under natural field conditions.

Dataset Statistics:

• Total images: 307 (2976×3968 pixels, RGB format)
• Total annotations: 1,516 bounding box labels
• Tobacco plant annotations: 984
• Weed annotations: 532
• Image format: JPG (72 dpi, 24-bit color depth)

Collection Details:

• Location: Tobacco fields in Erbaa, Tokat, Turkey
• Capture date: August 2023
• Camera height: Approximately 1.5 meters
• Capture angles: 60°-90°
• Lighting conditions: Natural daylight (morning hours)
• Equipment: Smartphone camera without flash

Annotation Process:

• Manual labeling using makesense.ai platform
• Expert validation and verification
• JSON format annotations compatible with object detection frameworks
• Stratified random sampling for dataset splits (80% train, 20% test)

Applications:

• Deep learning model training for agricultural object detection
• Precision agriculture research and development
• Weed detection and management systems
• Computer vision benchmarking in agricultural contexts

Data Organization: The dataset includes both raw images and corresponding annotation files. Train/test split information is provided for reproducible research.

Citation: If you use this dataset, please cite the associated research paper (currently under review).

Keywords: tobacco detection, weed detection, object detection, precision agriculture, computer vision, agricultural AI

**Access Note:**

This dataset is publicly available. If you use it in your research, please cite the following publication:

Y. Unal, "Impact of Attention Mechanisms and Focal Loss Tuning on RetinaNet Performance for Crop-Weed Detection: A Comparative Study With Anchor-Free Detectors," in IEEE Access, vol. 14, pp. 52626–52640, 2026, doi: 10.1109/ACCESS.2026.3680687

For questions or further information, contact: yunal@sinop.edu.tr