Panon
COMPUTER VISION, DEPLOYED

One platform
for every
detection

Upload a model, draw the region, set the confidence you trust. Panon runs it on the floor and shows you what it sees.

4 TARGETS TRACKED
ONNX · PYTORCH · YOLO
RTSP · GIGE · USB
EDGE GPU · ON-PREM
01 — LIVE METRICS

Confidence you
can measure.

LIVE|--:--:--
0
Frames inferred today
0.00%
Workstation uptime
0 ms
Mean edge latency
0
Cameras online
02 — HOW IT WORKS

Three steps from
model to running line.

01 / UPLOAD

Model and algorithm in, versioned

Drop an ONNX or PyTorch export. Panon fingerprints it, keeps every version and lets you roll back a line without redeploying hardware.

MODEL REGISTRY3 VERSIONS
weld-defect-v3.onnxACTIVE
weld-defect-v2.onnxARCHIVED
02 / CONFIGURE

Confidence, classes, region of interest

Draw the polygon that matters, name the classes you care about, and set a threshold per class. Everything outside the region is never inferred.

roi · zone-a
CONFIDENCE0.72
defectgap+ add class
03 / DEPLOY

Workstation and camera IP, live

Bind cameras to a workstation, push the config, watch inference start. Health, latency and drop rate report back per stream.

WS-03 / LINE-3RUNNING
cam-01 · 10.0.4.2130 fps
cam-02 · 10.0.4.2230 fps
+ bind camera—
03 — CAPABILITIES

Everything the line needs,
in one console.

04 — USE CASES

Where teams point it.

defect 0.94
MANUFACTURING QA

Catch the defect before the pallet

Weld, surface and assembly checks at line speed, with every rejection logged and reviewable.

count 1,284
LOGISTICS

Count, sort and read every label

Parcel counts, damage flags and OCR on the move, reconciled against the manifest.

no-ppe 0.88
WORKER SAFETY

Zones that alert, not cameras that record

PPE compliance and restricted-zone breaches, escalated the moment they happen.

RETAIL

Flow and dwell, without faces

Anonymous movement patterns per zone, aggregated into decisions you can act on.

PANON — EYE, IN SUNDANESE

Point it at anything.

Bring a model and a camera. We will have it deciding on your floor this week.