Rice Grain Inspection — Computer Vision Pipeline

Status: Executed notebook with preserved outputs • Last updated: 2025-08-15 04:39 PKT

This repository contains an end-to-end computer vision pipeline to detect, count, and measure rice grains from images. It uses Otsu thresholding, morphology, and connected-component analysis to isolate each grain, compute geometry (orientation, length, width), and export a CSV per folder. Rotated, cropped grain images can be saved for visual QA.

Highlights

• Batch processing across subfolders (each folder = variety/batch)
• Per-grain metrics: orientation (rad/deg), length, width (extensible to area, aspect ratio, solidity, etc.)
• Saves aligned ROI crops for inspection
• Cleaned notebook with ready outputs for portfolio/client review

Quick Results (from preserved notebook outputs)

• Images processed: 25
• Total grains detected (sum of per-image counts): 335

Per-folder totals:

• Adhwar: 60 grains across 5 images
• kernal: 75 grains across 5 images
• superIndianAdhwar: 69 grains across 5 images
• SuperIndianSaila: 52 grains across 5 images
• superkernalbasmatipurani: 79 grains across 5 images

Repository Structure

rice-cv-inspection/
├─ notebooks/
│  ├─ Rice_Analysis_Cleaned.ipynb      # with outputs, presentation-ready
│  ├─ Rice_Analysis_Lite.ipynb          # no outputs, tiny for quick preview
│  └─ AI Rice Image Analysis.ipynb      # original notebook (as provided)
├─ src/
│  └─ pipeline.py                        # modular skeleton (optional script path)
├─ data_sample/                          # put 5–10 example images here (optional)
├─ outputs_sample/                       # sample CSV/crops (optional)
├─ .gitignore
├─ LICENSE
├─ README.md
└─ requirements.txt

Getting Started

1. Clone this repository:

   git clone https://github.com/<your-username>/rice-cv-inspection.git
   cd rice-cv-inspection

2. (Optional) Create a virtual environment:

   python -m venv .venv
   source .venv/bin/activate   # Windows: .venv\Scripts\activate
   pip install -r requirements.txt

3. Open the notebook:

   • notebooks/Rice_Analysis_Cleaned.ipynb → already executed with outputs (no need to run).
   • notebooks/Rice_Analysis_Lite.ipynb → no outputs (fast to load on GitHub).

4. Run the pipeline later (optional):

   • Place your images under a root folder with subfolders per batch (or a single folder).
   • Adapt src/pipeline.py or the notebook cells to point to your path.
   • Outputs (CSV + crops) will be written under output/<folder>/ or a configured directory.

Tech Stack

• Python, NumPy, Matplotlib
• scikit-image (skimage) for thresholding, morphology, regionprops
• (Optional) scikit-learn for downstream classification/EDA

Roadmap

• Add EDA notebook: histograms of length/width/aspect ratio, scatter plots, correlation heatmaps
• Use regionprops major/minor axis for more geometry-robust measurements
• Small validation set to compare auto vs manual counts

License

This project is released under the MIT License (see LICENSE).

Credits

Maintained by Murtaza. Contributions and suggestions are welcome!