impulse-runner runs any Edge Impulse model exported as a Linux .eim file and
feeds it live data, with no code to write:
- camera models (classification, object detection, FOMO, visual anomaly)
get frames from a USB or CSI camera
- audio models (keyword spotting, sound classification) get the microphone
- IMU, environmental and sensor-fusion models get readings over a simple
HTTP API (POST /api/data), and any model can be called with POST /api/classify
Results are served as JSON and as a live web page, which Ubuntu Frame can
show full screen on an attached display.
Load a model by uploading it from the web page, by copying it to
/var/snap/impulse-runner/common/model.eim, or with
snap set impulse-runner model-url=https://.../model.eim
Options (snap set impulse-runner ...): port, input (auto|camera|microphone|api),
camera, camera-resolution, audio-device, audio-overlap, threshold, model-url,
upload-token.
Not affiliated with or endorsed by Edge Impulse.