📄 OMR Sheet Evaluation using OpenCV

This project implements an Optical Mark Recognition (OMR) system using Python and OpenCV. It evaluates a filled OMR sheet by comparing it against a reference (blank) OMR template and automatically calculates the final score with visual feedback.

---

🚀 Features

• Detects the main OMR region automatically

• Identifies and filters answer bubbles accurately

• Sorts bubbles column-wise and question-wise

• Detects marked answers using pixel density analysis

• Grades answers using a configurable answer key

• Visualizes:

  • ✅ Correct answers (Green)
  • ❌ Incorrect answers (Red)
  • ⚠️ Skipped questions (Orange)

• Displays final score directly on the evaluated image

---

📌 Technologies Used

• Python
• OpenCV
• NumPy
• Computer Vision techniques (Contours, Morphology, Thresholding)

---

🧠 How It Works (Pipeline)

1. Preprocessing

   • Converts image to grayscale

   • Applies Gaussian blur

   • Uses:

     • Adaptive Thresholding for reference image
     • OTSU Thresholding for filled image

2. OMR ROI (Region Of Interest) Detection

   • Finds the largest rectangular contour (main OMR container)
   • Crops the region of interest for accurate processing

3. Bubble Detection

   • Uses morphological operations

   • Filters contours based on:

     • Area
     • Aspect ratio
     • Circularity
     • Extent

4. Bubble Sorting

   • Sorts bubbles left-to-right (columns)
   • Then top-to-bottom (questions)

5. Answer Detection

   • Counts filled pixels inside each bubble
   • A bubble is considered marked if > 40% of its area is filled

6. Grading & Visualization

   • +2 marks for correct answers
   • −1 mark for incorrect answers
   • 0 for skipped questions
   • Draws color-coded circles on the evaluated OMR image

---

🧮 Scoring Logic

Case	Score
Correct	+2
Incorrect	−1
Skipped	0

---

🧩 Configuration

Currently configured for 20 questions:

block_of_questions = 4
questions_per_block = 5
options_per_question = 4

Answer Key Format

key = {
  1: "A", 2: "B", 3: "C", 4: "D",
  ...
}

You can easily modify this for different exams.

---

📂 Project Structure

.
├── images/
│   ├── OMR_ref2.png     # Reference (blank) OMR sheet
│   └── OMR_filled2.png  # Filled OMR sheet
├── main.py              # Main Python script
└── README.md

---

🛠️ Requirements

• Python 3.7+
• OpenCV
• NumPy

Install dependencies:

pip install opencv-python numpy

---

▶️ How to Run

python main.py

The program will:

• Display the graded OMR ROI with markings
• Print the final score in the console

---

🔄 Processing Flow

flowchart TD
    A[Reference OMR] --> B[Preprocessing]
    C[Filled OMR] --> B
    B --> D[Detect OMR ROI]
    D --> E[Bubble Detection]
    E --> F[Bubble Sorting]
    F --> G[Answer Detection]
    G --> H[Grading & Visualization]
    H --> I[Final Score]

Loading

---

📊 Output & Visual Results

This section demonstrates the end-to-end OMR evaluation pipeline, from the reference template to the final graded output.

---

🧾 Reference OMR Template

Used to detect the bubble layout and define the evaluation structure.

---

✍️ Filled OMR Sheet

Student-marked OMR sheet provided as input for evaluation.

---

🔢 Bubble Indexing Visualization

Illustrates how bubbles are detected, indexed, and mapped to questions and options internally.

---

✅ Final Graded Output

Color-coded evaluation with score overlay:

• 🟢 Green → Correct answers
• 🔴 Red → Incorrect selections
• 🟠 Orange → Skipped questions

---

🖥️ Console Output

Bubbles successfully filtered: 80
Detected 80 bubbles. Expected 80
Final Score: 32

---

🧩 Generate Your Own OMR Sheet (20 Questions)

To test or customize this project, you can generate your own 20-question OMR sheets using the online OMR generator below:

🔗 OMR Sheet Generator 👉 https://omrsheetgenerator.com/

📌 Recommended Settings

When generating the OMR sheet, use the following configuration for compatibility with this project:

• Number of Questions: 20
• Options per Question: 4 (A, B, C, D)
• Layout: 4 columns × 5 questions per column
• Bubble Shape: Circle
• Print Scale: 100% (no scaling)

⚠️ Important: Any change in layout (number of questions, columns, or options) requires updating the configuration values in the code:

block_of_questions
questions_per_block
options_per_question

---

🖨️ Usage Tip

• Use the generated sheet as:

  • Reference Image → unfilled OMR
  • Filled Image → marked OMR

• Ensure:

  • Good lighting
  • Minimal skew
  • Clear bubble filling (pen/pencil)

• Use virtual environment

---

⚠️ Notes & Limitations

• Assumes consistent OMR layout
• Bubble count must exactly match configuration
• Lighting conditions should be reasonably uniform
• Currently supports only one OMR format at a time

---

🔮 Possible Enhancements

• 📄 Support multiple OMR templates
• 📦 Automatic question count detection
• 📊 CSV / Excel export
• 🖥️ GUI interface
• 🤖 ML-based bubble classification

---