plant_care

🥦 Cauliflower Disease Detection — YOLOv8 Model

📌 Overview

This repository hosts a trained YOLOv8 object detection model for identifying cauliflower leaf diseases and stress conditions from images.

The model was trained on a curated, multi-class agricultural dataset and is suitable for:

• Automated crop disease detection
• Smart agriculture & decision support systems
• Mobile and edge inference
• Research in plant pathology & computer vision

The model detects five distinct disease categories from field images and outputs bounding boxes and class labels.

---

🧠 Model Summary

---

🏷️ Supported Classes

---

⚙️ Training Configuration

The model was trained using Ultralytics YOLOv8. The exact training arguments used are listed below:

# Key Training Parameters
task: detect
model: yolov8s.pt
data: cauli_disease-3/data.yaml

# Training Setup
epochs: 200
batch: 32
imgsz: 640
optimizer: auto
pretrained: true
amp: true
seed: 0
deterministic: true

# Optimization
lr0: 0.01
momentum: 0.937
weight_decay: 0.0005
warmup_epochs: 3.0

# Validation & Inference
iou: 0.7
max_det: 300
split: val

# Augmentation
mosaic: 1.0
fliplr: 0.5
auto_augment: randaugment

# Tracking & Output
tracker: botsort.yaml
save_dir: runs/detect/train2

Training outputs and intermediate artifacts were saved under runs/detect/train2/.

---

📊 Dataset

• Dataset: Cauliflower Disease Detection Dataset
• Annotation Format: YOLO
• Classes: 5 (see Supported Classes)
• License: CC BY 4.0

The dataset contains real-field cauliflower leaf images annotated with bounding boxes for disease localization. The dataset used during training is referenced in data above (cauli_disease-3/data.yaml).

---

🚀 Inference Usage

Using Ultralytics CLI

yolo detect predict \
  model=weights/best.pt \
  source=your_image.jpg \
  imgsz=640

Python Inference Example

from ultralytics import YOLO

model = YOLO("weights/best.pt")
results = model("image.jpg", conf=0.25)

# Display or save results
results[0].show()
results[0].save(save_dir='inference_results')

Notes:

• Set conf (confidence) threshold to trade off precision vs. recall.
• For batch or directory inference, pass a folder path to source.

---

📦 Repository Contents (suggested)

.
├── best.pt                # Best trained weights (model for inference)
├── last.pt                # Final checkpoint from training
├── args.yaml              # Training configuration 
├── results.png            # Training curves / metrics plot
├── README.md              # This documentation file

---

📈 Performance Notes

• Trained with data augmentation options enabled during training.
• Early stopping monitored with patience = 100.
• IoU threshold set to 0.7 during validation.

Exact evaluation metrics ([email protected], precision, recall) should be calculated on a held-out test set and can be added to this README once available.

---

🌍 Deployment Use Cases

• Smart farming dashboards
• Mobile crop monitoring applications
• Drone-based field inspection
• Agricultural advisory systems
• Edge AI inference pipelines (ONNX / TensorRT conversions possible)

---

📜 License

This model and its training outputs are released under the MIT license. Attribution is required for use in publications or products.

---

📖 Citation

If you use this model in academic or applied work, please cite the model and dataset:

@model{cauliflower_yolov8_detector,
  title={YOLO based Cauliflower Disease Detection Model},
  author={Indra Prasad Sapkota},
  year={2025},
  url={https://huggingface.co/Bishal17/plant_care}
}

---

🤝 Acknowledgements

• Model trained using Ultralytics YOLOv8.
• Dataset annotation and preprocessing assisted by Roboflow or similar annotation tools.
• Model hosting by Hugging Face (recommended).

---

📬 Contact & Contributions

Contributions are welcome (issues / PRs). Possible contributions:

• Performance benchmarks and evaluation reports
• Edge optimization (quantized, ONNX, TensorRT)
• Additional datasets and annotated examples

For contributions or questions, open an issue on the repository or contact the maintainer listed on the Hugging Face project page.