Skip to content

Feat: Implement Image Metadata Extraction Pipeline

Kushal Lagichetty requested to merge feat/img-pipeline into feat/multi-format-pipeline-phase1

Summary

This MR introduces a specialized extraction pipeline for image formats (.jpg, .jpeg, .png). It allows the system to extract embedded EXIF data, GPS coordinates, and technical image properties without requiring OCR or LLM processing, significantly improving performance for image-only metadata tasks.

Key Changes

1. Metadata Extraction Logic

  • bookextractor/image_utils.py: Added extract_image_metadata using Pillow and piexif to parse image headers.
  • GPS Normalization: Implemented get_decimal_from_dms to convert Degrees/Minutes/Seconds (DMS) rational tuples into standard decimal degrees.
  • Field Coverage: Added extraction for camera make/model, lens info, date taken, DPI, bit depth, and color space.

2. Architectural Updates

  • Models: Added the ImageMetadata Pydantic model to bookextractor/models.py.
  • Pipeline Router: Updated ExtractionPipeline in bookextractor/pipeline.py with a process_image method.
  • Unified Routing: Updated main.py (FastAPI and Typer CLI) to automatically route files to the image pipeline based on extension.

3. Engineering & Tooling

  • Dependencies: Added piexif to pyproject.toml and updated uv.lock.
  • Validation: Added tests/test_image_utils.py with unit tests for GPS normalization and EXIF extraction.
  • Linting: Resolved multiple Ruff, Mypy, and pre-commit hook issues across modified files.

How to Test

CLI

bookextractor path/to/image.jpg output.json

API

# 1. Start the server
bookextractor --api

# 2. Send a POST request to /extract with an image file
curl -X POST -F "[email protected]" http://localhost:8000/extract

Automated Tests

uv run pytest tests/test_image_utils.py

Example Output

{
  "width": 4032,
  "height": 3024,
  "format": "JPEG",
  "dpi_horizontal": 72.0,
  "dpi_vertical": 72.0,
  "color_space": "RGB",
  "bit_depth": 24,
  "exif_camera_make": "Apple",
  "exif_camera_model": "iPhone 13",
  "exif_date_taken": "2023:10:27 10:00:00",
  "exif_gps_latitude": 34.053055,
  "exif_gps_longitude": -118.245833,
  "exif_lens": "iPhone 13 back dual wide camera 5.1mm f/1.6"
}

Checklist

  • New dependency piexif added.
  • GPS normalization handles N/S/E/W references.
  • API and CLI both support the new extensions.
  • Tests pass and coverage is maintained.
  • Linting and type checking (Ruff/Mypy) are clean.

Merge request reports