gaitkit 1.4.4

Universal gait event detection toolkit — 10 methods including BIKE, with C-accelerated backends

gaitkit (Python)

Python package for gait event detection from motion-capture data.

Install

Current status: PyPI release available (latest stable on PyPI).

Install from local source:

python -m pip install -e ./python
python -m pip install -e "./python[all]"   # optional extras

Once published:

pip install gaitkit

Optional extras:

pip install "gaitkit[all]"   # onnx + deep + viz

Quick Start

import gaitkit

trial = gaitkit.load_example("healthy")
result = gaitkit.detect(trial, method="bike")
print(result.summary())

# Optional: combine C3D markers with an external angle file
result2 = gaitkit.detect("trial_07.c3d", method="bike", angles="res_angles_t.mat")

DeepEvent weights are downloaded automatically on first DeepEvent use and cached in ~/.cache/gaitkit/.

Proprietary JSON I/O (MyoGait-like)

The compatibility API accepts a proprietary JSON payload with angles.frames and can export back a MyoGait-compatible events JSON.

import gaitkit

out = gaitkit.detect_events_structured("bike", "myogait_output_no_events.json", fps=100.0)
paths = gaitkit.export_detection(out, "outputs/trial_07", formats=("json", "myogait"))
print(paths)

Input fields recognized per frame:

• frame_idx
• hip_L, knee_L, ankle_L
• hip_R, knee_R, ankle_R
• pelvis_tilt, trunk_angle
• optional landmark_positions

MyoGait-compatible output file (*_myogait_events.json) contains:

• events.method
• events.fps
• events.left_hs, events.right_hs, events.left_to, events.right_to as arrays of {frame, time, confidence}.

Project

• Repository: https://github.com/IDMDataHub/gaitkit
• Issue tracker: https://github.com/IDMDataHub/gaitkit/issues
• Reproducibility: https://github.com/IDMDataHub/gaitkit/blob/master/REPRODUCIBILITY.md

Development testing

python -m pip install -e ./python
python -m unittest -v

Build distributions

python -m pip install build
cd python
python -m build