Skip to main content

Version 3.3.5 Release Notes

Release Date: 2025-11-01
Type: Maintenance Release

📦 Overview​

Version 3.3.5 is a maintenance release that updates version references across all configuration files and documentation. This release maintains full compatibility with v3.3.4 and includes all critical fixes from previous releases.

🔄 Changes​

Version Updates​

  • Updated version to 3.3.5 in all configuration files:
    • pyproject.toml
    • ign_lidar/__init__.py
    • docs/package.json
    • conda-recipe/meta.yaml
    • docs/docusaurus.config.ts
  • Updated documentation references:
    • README.md
    • docs/docs/intro.md
    • CHANGELOG.md

✨ Included Features​

This release includes all features and fixes from previous versions:

From v3.3.4 (Critical Bug Fix)​

  • 🔴 CRITICAL: Fixed BD TOPO reclassification priority (+20-30% building classification accuracy)
  • ✨ NEW: Unified feature filtering for planarity, linearity, and horizontality
  • 95% artifact reduction in geometric features
  • 100% elimination of NaN/Inf warnings

From v3.3.3 (Performance Improvements)​

  • 10× faster DTM lookup with RTM spatial indexing
  • Intelligent gap filling for missing DTM values
  • Automatic memory optimization prevents OOM crashes
  • 40-50% faster processing with memory-optimized configuration
  • +30-40% facade detection improvement
  • Building cluster IDs for instance segmentation

From v3.1.0 (Unified Feature Filtering)​

  • Generic filtering API for any geometric feature
  • Specialized functions for planarity, linearity, horizontality
  • Adaptive spatial filtering with variance detection
  • ~60% code reduction through unified implementation

🔄 Migration Guide​

From v3.3.4 to v3.3.5​

Required Actions:

  1. Upgrade package:

    pip install --upgrade ign-lidar-hd

  2. Verify version:

    ign-lidar-hd --version
    # Should show: ign-lidar-hd 3.3.5

  3. Or via Python:

    import ign_lidar
    print(ign_lidar.__version__)
    # Should show: 3.3.5

Breaking Changes​

None! This release is 100% backward compatible with v3.3.4.

📊 Performance & Quality​

All performance metrics from v3.3.4 are preserved:

Classification Accuracy​

Metricv3.3.5
Building classification (BD TOPO)94-97%
Overall classification rate94-97%

Feature Quality​

FeatureArtifactsStatus
Planarity5-10/tile✅
Linearity3-8/tile✅
Horizontality2-6/tile✅
NaN/Inf warningsEliminated✅

🔧 System Requirements​

Minimum (CPU Only)​

  • Python 3.8+
  • 16GB RAM (with memory-optimized config)
  • 20GB disk space
  • Python 3.8+
  • 28-32GB RAM (memory-optimized) or 64GB+ (full quality)
  • NVIDIA GPU with 12-14GB VRAM (RTX 3060/3070) or 16GB+ (RTX 4080/4090)
  • CUDA 11.8+ or 12.x
  • 50GB disk space

📦 Dependencies​

No dependency changes from v3.3.4.

Core dependencies:

  • numpy>=1.21.0
  • laspy>=2.3.0
  • scikit-learn>=1.0.0
  • scipy>=1.7.0
  • numba>=0.56.0
  • hydra-core>=1.3.0
  • omegaconf>=2.3.0

Optional (GPU):

  • cupy-cuda11x or cupy-cuda12x
  • cuml (RAPIDS)
  • cuspatial (RAPIDS)

📖 Related Releases​

  • v3.3.4 - Critical bug fix + unified filtering
  • v3.3.3 - Performance optimizations and DTM improvements
  • v3.2.1 - Rules framework and configuration enhancements
  • v3.1.0 - Unified feature filtering framework
  • v3.0.6 - Planarity artifact filtering

📚 Documentation​

🤝 Support​

Made with ❤️ for the LiDAR and Machine Learning communities