vision-calibration-pipeline

End-to-end calibration workflows and a session API for calibration-rs.

This crate provides two complementary approaches for camera calibration:

1. Session API: Structured workflows with artifact management + JSON checkpointing
2. Imperative Functions: Direct access to pipeline functions for custom workflows

Features

• Planar intrinsics calibration: Zhang's method with Brown-Conrady distortion
• JSON I/O: Serialize/deserialize inputs, configs, and results
• Checkpointing: Save and resume session state

Dual API Design

Session API

Best for standard workflows with branching, artifact tracking, and checkpointing:

use vision_calibration_pipeline::planar_intrinsics::{PlanarIntrinsicsConfig, PlanarIntrinsicsProblem};
use vision_calibration_pipeline::session::{CalibrationSession, ExportOptions, FilterOptions};
use vision_calibration_pipeline::PlanarDataset;

// Create session
let mut session = CalibrationSession::<PlanarIntrinsicsProblem>::new();

// Add observations (a PlanarDataset: Vec<View<NoMeta>>)
let obs_id = session.add_observations(dataset);

// Initialize (linear solver)
let config = PlanarIntrinsicsConfig::default();
let init_id = session.run_init(obs_id, config.clone())?;

// Save checkpoint
let checkpoint = session.to_json()?;

// Optimize (non-linear refinement)
let result_id = session.run_optimize(obs_id, init_id, config)?;

// Filter outliers and re-optimize (optional)
let obs_filtered = session.run_filter_obs(obs_id, result_id, FilterOptions::default())?;

// Export results
let estimate = session.run_export(result_id, ExportOptions::default())?;

Imperative Functions API

Best for custom workflows requiring intermediate inspection:

use vision_calibration_pipeline::planar_intrinsics::{
    planar_init_seed_from_views, run_planar_intrinsics, PlanarIntrinsicsConfig,
};
use vision_calibration_pipeline::PlanarDataset;

let config = PlanarIntrinsicsConfig::default();
let seed = planar_init_seed_from_views(&dataset, config.init_opts.clone())?;
let estimate = run_planar_intrinsics(&dataset, &config)?;

JSON I/O Example

use vision_calibration_pipeline::{PlanarDataset, PlanarIntrinsicsConfig, run_planar_intrinsics};

// Load from JSON
let input: PlanarDataset = serde_json::from_str(&input_json)?;
let config: PlanarIntrinsicsConfig = serde_json::from_str(&config_json)?;

// Run calibration
let estimate = run_planar_intrinsics(&input, &config)?;

// Save results
let output_json = serde_json::to_string_pretty(&estimate)?;

When to Use Each API

See Also

• vision-calibration-core: Math types and camera models
• vision-calibration-linear: Linear initialization solvers
• vision-calibration-optim: Non-linear optimization
• Book: Pipelines
• functions.md: Comprehensive API reference