Abstract
This paper presents COT-AD, a comprehensive Dataset designed to enhance cotton crop analysis through computer vision. Comprising over 25,000 images captured throughout the cotton growth cycle, with 5,000 annotated images, COT-AD includes aerial imagery for field-scale detection and segmentation and high-resolution DSLR images documenting key diseases. The annotations cover pest and disease recognition, vegetation, and weed analysis, addressing a critical gap in cotton-specific agricultural datasets. COT-AD supports tasks such as classification, segmentation, image restoration, enhancement, deep generative model-based cotton crop synthesis and early disease management, advancing data-driven crop management.
Field Views and Data Examples

Results : Image enhancement and Sementic segmentation


About COT-AD Dataset
The COT-AD dataset supports research in agricultural technology, especially for cotton crop monitoring, disease detection, Computer Vision, Image Processing, and precision farming. It consists of two main sections: Annotated Data and Unannotated Data, each tailored for different crop analysis and model development needs.
Annotated Data
Aerial Images for Detection and Segmentation
This section includes images collected over six months, divided into four parts:
- Part A: First two months
- Part B: Third month
- Part C: Fourth month
- Part D: Fifth and sixth months
Each part contains four subfolders:
- Images: Aerial JPG images of cotton crops
- Detection Labels: YOLO-formatted .txt files (single-class detection)
- Segmentation Masks: Binary JPG masks for segmentation
- Segmentation Labels: YOLO-formatted .txt files for segmentation
High-Resolution DSLR Images for Cotton Crop Insights
This section focuses on different aspects of cotton crop health:
- Leaf Disease:
- Yellowish Leaf
- Leaf Spot Bacterial Blight
- Leaf Reddening
- Fresh Leaf
- Cotton Boll:
- Boll Rot
- Damaged Cotton Boll
- Healthy Cotton Boll
- Bugs:
- Mealy Bug
- Red Cotton Bug
Unannotated Data
DSLR Data (Week-wise)
High-resolution close-up images of cotton crops, capturing detailed views of leaves, bolls, and bugs. Includes videos for temporal analysis. Ideal for disease detection, pest monitoring, and growth stage identification.
Farm_2 (Biweekly Image Acquisition)
Drone images captured biweekly at three altitudes (10m, 15m, and 115m):
- 10m and 15m: Provide detailed crop views
- 115m: Offers a wide field overview
Enables research on plant development, disease detection, and environmental monitoring across different growth stages.
Farm_1 (Weekly Image Acquisition)
Weekly drone images at altitudes of 10m, 15m, and 115m. Facilitates time-series analysis, crop growth tracking, and anomaly detection through consistent, high-frequency monitoring.
Explore the Dataset
Explore the dataset and contribute to the future of agricultural technology!
Citations
coming soon
@misc{ali2025cotad, title={COT-AD: Cotton Analysis Dataset}, author={Akbar Ali and Mahek Vyas and Soumyaratna Debnath and Chanda Grover Kamra and Jaidev Sanjay Khalane and Reuben Shibu Devanesan and Indra Deep Mastan and Subramanian Sankaranarayanan and Pankaj Khanna and Shanmuganathan Raman}, year={2025}, eprint={2507.18532}, archivePrefix={arXiv}, primaryClass={cs.CV} url = {http://arxiv.org/abs/2507.18532}, }