media-analyzer 0.3.0

Analyze video/image with machine learning methods, exif data, and other file based information.

Media Analyzer

Media Analyzer is a Python library designed to analyze media files, providing insights into their content and metadata. It supports various functionalities, including image classification, captioning, optical character recognition (OCR), and facial recognition.

Features

• GPS: Gather GPS coordinates from exif, and reverse geocode to get the country, province and city.
• Exif: Extract exif data/metadata from photos, videos, gifs, etc.
• Weather: Get weather info at the time the photo/video was taken.
  • temperature
  • dewpoint
  • relative humidity
  • precipitation
  • wind gust
  • pressure
  • sun hours
  • condition

• Quality Detection: Detect objectively measurable quality of images:
  • Sharpness
  • Noise
  • Exposure
  • Dynamic range
  • Color clipping (white/black levels)
  • A composite score is generated, so photos can be ranked by quality.

• Image Classification: Identify objects, scene type, activities, animals, and events present in images.
• Image Captioning: Generate descriptive captions for images using models like BLIP or LLM-based captioners.
• Embedding: Generate clip embeddings for images and text. Can be used to cluster images or search semantically through images.
• Optical Character Recognition (OCR): Extract text from images to identify documents, receipts, menus, and more.
• LLM: Get detailed image summary, more indepth than just a caption, using an LLM. Can also be used to generate a summary of a document shown in a photo or video.
• Facial Recognition: Detect faces in images and provide details such as age, sex, bounding box, and facial landmarks. Includes an embedding of the face, which can be used for clustering.
• Datetime Taken: Photo and video files are messy and have unreliable datetime tags. This packages uses six different methods with varying priority to get the datetime a photo is taken, including the timezone if possible.
• Data Url: Generate data url for tiny preload thumbnail.

Installation

To install Media Analyzer, use pip:

pip install media-analyzer

Requirements

You must have the following in PATH.

• ExifTool: https://exiftool.org/
• Tesseract OCR: https://tesseract-ocr.github.io/tessdoc/Installation.html

Examples

Example output of the main analyze function can be viewed at example_output.json. Further example code is available at /examples.

Usage

Here's a basic example of how to use Media Analyzer:

from media_analyzer import MediaAnalyzer
from pathlib import Path

analyzer = MediaAnalyzer()
media_file = Path("image.jpg")
result = analyzer.photo(media_file)

print(result)

Disable analysis modules

The analysis is done based on modules, the following modules are available and enabled by default (click the links to see the possible modules).

File-based Modules

Visual Modules

Modules can be turned off by changing the config provided to the MediaAnalyzer class:

from media_analyzer import MediaAnalyzer, AnalyzerSettings, FileModule, VisualModule
from pathlib import Path

config = AnalyzerSettings(
    enabled_file_modules={FileModule.EXIF},  # Only do exif data analysis on file
    enabled_visual_modules={VisualModule.CAPTION},  # Only do caption module as visual module
)
analyzer = MediaAnalyzer(config=config)
media_file = Path(__file__).parents[1] / "tests/assets/tent.jpg"
result = analyzer.photo(media_file)

Configuration

The AnalyzerSettings class allows you to customize various aspects of the analysis:

media_languages: List of languages for OCR to consider.
theme_color_variant: The color variant used for the generated theme.
captions_provider: The provider for image captioning (e.g., 'BLIP', 'LLM').
llm_provider: The provider for the large language model (LLM), which can be used for summaries and captions.
enable_text_summary: Enable or disable text summarization.
enable_document_summary: Enable or disable document summarization.
document_detection_threshold: Confidence threshold for document detection.
face_detection_threshold: Confidence threshold for face detection.
enabled_file_modules: List of file modules to enable (e.g., exif data, gps, weather detection).
enabled_visual_modules: List of visual modules to enable (e.g., 'classification', 'captioning', 'ocr', 'facial_recognition').

Full docs can be found at https://ruurdbijlsma.github.io/media-analyzer.

Attribution

• Meteostat for weather info: https://dev.meteostat.net/python/
• Reverse geocoding data from geonames: https://download.geonames.org/
• ExifTool for exif and similar data: https://exiftool.org/