ExamReader

Exam grading tool for .NET — OCR-powered answer sheet processing, automatic grading, and analytics for educators.

Why ExamReader?

Grading 100 multiple-choice exams by hand takes hours and is error-prone. ExamReader processes a stack of scanned answer sheets in seconds — with per-question analytics that reveal which topics your class struggles with.

Demo — Try Without Setup

git clone https://github.com/mcandiri/ExamReader.git
cd ExamReader
dotnet run --project src/ExamReader.Web
# Open http://localhost:5000 → Click "Demo Mode"

Pre-loaded with 25 students and 30 questions. See the full dashboard, analytics, and export features — no API keys or database needed.

Features

Multi-Provider OCR

Provider	Cost	Speed	Setup
Azure Computer Vision	Pay-per-use	Fast	API key
Tesseract	Free	Medium	Local install
Demo Mode	Free	Instant	None

Answer Sheet Support

• Bubble sheets (A/B/C/D/E)
• Grid-based answer sheets
• Written answer extraction

Smart Grading

• Configurable negative marking (-0.25 per wrong answer)
• Weighted questions (different point values)
• Partial credit for multi-select
• Custom letter grade scales (A/B/C/D/F with +/- variants)

Class Analytics

• Score distribution histogram
• Per-question difficulty & discrimination index
• Automatic flagging of "bad questions" (low discrimination)
• Most common wrong answers per question

Batch Processing

• Process entire classroom at once
• Real-time progress tracking
• Handles blank answers, multiple marks, smudges

Export

• HTML report with charts (standalone, share via email)
• JSON for integration with other systems
• CSV for Excel/Google Sheets

Limitations

• OCR accuracy depends on scan quality — blurry or skewed images will produce unreliable results. Clean, high-contrast scans work best.

• Demo mode uses synthetic data — the 25-student dataset is generated to show realistic score distributions, not captured from real exams.

• Azure and Tesseract providers require setup — Demo mode is the only zero-config option.

• No OMR (Optical Mark Recognition) — current implementation uses text-based OCR, not dedicated bubble detection. OMR is on the roadmap.

• In-memory only — no database persistence. Results are available during the session and via export.

Architecture

┌─────────────────────────────────────────────────────┐
│                  Blazor Server UI                     │
│  Home │ Exam Setup │ Processing │ Results │ Demo     │
├─────────────────────────────────────────────────────┤
│                  ExamReader.Core                      │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐          │
│  │ OCR      │  │ Grading  │  │ Analytics│          │
│  │ Provider │→ │ Engine   │→ │ Engine   │          │
│  └──────────┘  └──────────┘  └──────────┘          │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐          │
│  │ Parsing  │  │ Batch    │  │ Reports  │          │
│  │ Engine   │  │ Processor│  │ Generator│          │
│  └──────────┘  └──────────┘  └──────────┘          │
├─────────────────────────────────────────────────────┤
│  OCR Providers: Azure CV │ Tesseract │ Demo          │
└─────────────────────────────────────────────────────┘

Project Structure

ExamReader/
├── src/
│   ├── ExamReader.Core/          # Core library
│   │   ├── Models/               # Domain models
│   │   ├── Ocr/                  # OCR providers
│   │   ├── Parsing/              # Answer sheet parsers
│   │   ├── Grading/              # Grading engine
│   │   ├── Batch/                # Batch processing
│   │   ├── Analytics/            # Exam analytics
│   │   ├── Reports/              # Report generators
│   │   └── Demo/                 # Demo data
│   └── ExamReader.Web/           # Blazor Server app
│       ├── Components/Pages/     # Blazor pages
│       ├── Shared/               # Shared components
│       └── Services/             # Web services
├── tests/
│   ├── ExamReader.Core.Tests/    # Core library tests
│   └── ExamReader.Web.Tests/     # Web app tests
├── .github/workflows/ci.yml     # CI pipeline
└── ExamReader.sln

Configuration

{
  "OcrProvider": "Demo",
  "Azure": {
    "ComputerVision": {
      "Endpoint": "https://your-resource.cognitiveservices.azure.com/",
      "ApiKey": "your-api-key"
    }
  }
}

Born From Production

Started as a quick Azure Computer Vision experiment to automate multiple-choice grading. Over time, teachers asked for deeper insights — "which questions are too easy?", "which ones don't discriminate between good and weak students?". The analytics module (difficulty index, discrimination index, distractor analysis) was added to answer those questions.

Security

• Scanned images processed in memory, never persisted to disk
• API keys stored in configuration only, never logged
• No student data sent to external services in demo mode

Roadmap

• OMR (Optical Mark Recognition) with OpenCV for higher accuracy
• PDF answer sheet support
• Multi-page answer sheet support
• Integration with LMS platforms (Moodle, Canvas)
• Mobile app for scanning with phone camera
• AI-powered written answer grading (essay scoring)

License

MIT