epica 0.1.4

Trajectory alignment and evaluation toolkit

EPIC-Alignment (epica)

epica is a trajectory alignment and evaluation toolkit.

It provides:

• 3-step alignment pipeline (time offset, extrinsic, world alignment)
• a set of CLI tools (traj, ape, rpe, res, config)
• OpenVINS compatibility entrypoints
• optional plotting and rerun-based visualization

Installation

Create and activate a virtual environment first (recommended):

conda create -n epa python=3.10 -y
conda activate epa

Then install:

pip install epica

Optional extras:

pip install "epica[rerun]"  # rerun visualization
pip install "epica[ros]"    # bag / bag2 / mcap support
pip install "epica[geo]"    # map-related tools

Quick Start

For General Workspace:

epa --gt-csv <gt_file> --gt-format <gt_format> --est-path <est_file> --est-format <est_format> --plot

OpenVINS examples:

Single case:

epa_openvins /path/to/case_dir se3 --keep-output

Multiple cases:

epa_openvins /path/to/case_a /path/to/case_b --align-mode se3

Run from OpenVINS repo root:

epa_openvins ./ov_eval/example none --keep-output

Each case directory should contain:

• stamped_groundtruth.txt
• stamped_traj_estimate.txt

Outputs

Each run generates a run folder with:

• plots/
• metrics.json
• metrics_summary.csv
• report_en.md
• report_zh.md

Benchmark LaTeX Tables

For benchmark summary CSV files, generate paper tables with:

epa_latex_summary --summary-csv /path/to/summary.csv

This creates:

• main_table.tex
• dataset_table.tex
• appendix_full_table.tex

Full CLI Toolchain

• epa / epica / epic-alignment: Run the main 3-step EPA pipeline for alignment and metric export; usage: epa --gt-csv gt.tum --gt-format tum --est-path est.tum --est-format tum --plot.
• epa_config: Manage global or per-tool default config values; usage: epa_config set --tool epa_ape plot_mode xy.
• epa_ape: Compute APE in EVO-style format with optional alignment and plots; usage: epa_ape tum gt.tum est.tum --align --plot.
• epa_rpe: Compute RPE in EVO-style format with configurable delta settings; usage: epa_rpe tum gt.tum est.tum --delta 1 --delta_unit f --plot.
• epa_traj: Compare, align, sync, and visualize trajectories in EVO-style workflow; usage: epa_traj tum gt.tum est.tum --sync --align --plot.
• epa_fig: Re-render saved plotting specs without recomputing metrics; usage: epa_fig outputs/ape_plot.json --save_plot outputs/ape.png.
• epa_res: Compare multiple result bundles or EVO zip results; usage: epa_res run_a.zip run_b.zip --metric all --plot.
• epa_benchmark: Run batch benchmark harness for EPA/EVO-style comparison over many cases; usage: epa_benchmark --cases-csv cases.csv --out-dir outputs/bench.
• epa_plot_summary: Generate benchmark summary figures from harness outputs; usage: epa_plot_summary --summary-csv outputs/bench/run_xxx/summary.csv.
• epa_latex_summary: Generate paper-ready LaTeX tables from benchmark summary.csv; usage: epa_latex_summary --summary-csv outputs/bench/run_xxx/summary.csv.
• epa_metric_res: Aggregate or compare one or more metrics.json files directly; usage: epa_metric_res --inputs run1/metrics.json run2/metrics.json.
• epa_case_rerun / epa_rerun: Replay one case with rerun-based visual diagnostics; usage: epa_case_rerun --case-json outputs/bench/run_xxx/cases_json/case_001.json.
• epa_ipython: Launch EPA tools in IPython-friendly entry mode; usage: epa_ipython.
• epa_ov_eval: OpenVINS ov_eval compatibility umbrella entrypoint; usage: epa_ov_eval --help.
• epa_ov_format_converter: Convert trajectory formats in ov_eval-compatible style; usage: epa_ov_format_converter --help.
• epa_ov_plot_trajectories: Plot trajectories in ov_eval-compatible style; usage: epa_ov_plot_trajectories --help.
• epa_ov_error_singlerun: Run single-run error evaluation in ov_eval-compatible style; usage: epa_ov_error_singlerun --help.
• epa_ov_error_dataset: Run dataset-level error summary in ov_eval-compatible style; usage: epa_ov_error_dataset --help.
• epa_ov_error_comparison: Run algorithm comparison summary in ov_eval-compatible style; usage: epa_ov_error_comparison --help.
• epa_openvins: Run EPA on one or multiple OpenVINS case folders with minimal integration flow; usage: epa_openvins /path/to/case_dir se3 --keep-output.

Project Links

• Source and full project docs: Docs