Can Modern C++ Speed Up Your Bundle Adjustment Pipeline?

Explore how modern C++23 features can revolutionize Bundle Adjustment pipelines for 3D vision applications in this comprehensive 58-minute conference talk from CppCon 2025. Learn to build robust, efficient, and extensible Bundle Adjustment systems capable of handling large-scale optimization problems commonly found in structure-from-motion and SLAM applications. Discover practical implementation strategies using Ceres Solver combined with quaternion-based pose parameterization and manifold support to ensure stable rotation handling and numerical consistency. Master outlier rejection techniques through RANSAC filtering and reprojection error validation, while implementing spatial smoothing to eliminate local geometric inconsistencies using clean, expressive code with std::ranges and views::filter. Dive deep into Schur complement optimization tricks for grouping cameras and points separately to accelerate computation performance. Compare full Bundle Adjustment approaches against sliding window strategies and understand when to apply preprocessing filters versus robust loss functions like Huber for handling noisy observations. Gain insights into scaling Bundle Adjustment to real-world datasets like BAL (Bundle Adjustment in the Large) while addressing common challenges including numerical instability and pipeline maintainability. Learn practical techniques for building scalable and reliable computer vision pipelines that leverage the latest C++23 features for complex geometric optimization tasks in warehouse automation and robotics applications.