Skip to content

CHANGELOG.md

Latest commit

 

History

History
71 lines (57 loc) · 3.2 KB

CHANGELOG.md

File metadata and controls

71 lines (57 loc) · 3.2 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.2.0] - 2026-02-22

Changed

  • Flexible OpenCV dependencies - OpenCV is no longer a hard dependency
    • Users can now choose their preferred OpenCV variant:
      • pip install brisque[opencv-python] - Desktop applications with GUI
      • pip install brisque[opencv-python-headless] - Servers/Docker (recommended)
      • pip install brisque[opencv-contrib-python] - Desktop with extra modules
      • pip install brisque[opencv-contrib-python-headless] - Servers with extra modules
    • Users with existing OpenCV installations can install brisque without conflicts
    • Helpful error message when OpenCV is missing with installation instructions

Removed

  • requirements.txt - Replaced by requirements-dev.txt for development only
    • Users should rely on setup.py for package installation
    • requirements-dev.txt contains only development dependencies (pytest, black, etc.)

Fixed

  • Improved build configuration in pyproject.toml (removed runtime dependencies from build-system)
  • Updated minimum Python version requirement to 3.7+
  • Updated classifiers to reflect current Python version support (3.7-3.12)

Migration Guide

  • From v0.1.x: No code changes required. Your existing code will work.
    • If upgrading: pip install --upgrade brisque (your OpenCV will be detected)
    • For new installs: pip install brisque[opencv-python-headless] (or your preferred variant)

[0.1.0] - 2026-02-22

Added

  • BRISQUETrainer class - Complete training workflow for custom BRISQUE models
    • add_image(image, score) - Add single image with quality score
    • add_dataset(image_dir, scores_file) - Load images and scores from CSV
    • train(svm_c, svm_gamma, svm_epsilon, kernel) - Train SVM regression model
    • save_model(svm_path, norm_path) - Save trained model for later use
    • evaluate(images, scores) - Evaluate model performance (RMSE, PLCC, SROCC)
    • clear() - Reset trainer state
  • Custom model support - BRISQUE class now accepts custom trained models
    • model_path parameter in BRISQUE constructor for custom model files
  • Comprehensive test suite - 56 tests covering all functionality
  • Documentation - Docstrings for all public methods and classes

Fixed

  • scipy compatibility - Added workaround for libsvm's use of deprecated scipy.ndarray
    • Works with scipy 1.8+ which removed scipy.ndarray
  • Feature extraction - Fixed array conversion issues in trainer module
  • Image handling - Better support for various image formats:
    • RGBA images (alpha channel removed)
    • Grayscale images
    • Single-channel images with 3D shape (H, W, 1)
    • Float images normalized to [0, 1]

Changed

  • Version bumped from 0.0.18 to 0.1.0
  • Improved error messages for invalid inputs

Dependencies

  • No changes to dependencies - uses same libsvm, scipy, numpy, opencv-python, scikit-image

[0.0.18] - Previous Release

Note

  • Original BRISQUE implementation with pre-trained model
  • Basic quality scoring functionality