Workflows

Run vision pipelines over images and video.

Build a pipeline in a visual editor, then run it over an image, a video, or a dataset. Track objects, count crossings, and measure dwell time.

workflow.py
from pictograph import Client

client = Client()

# Run a workflow over a video
run = client.workflows.run(
    workflow_id="wf_door_counter",
    source={"type": "video", "url": "./entrance.mp4"},
)

run = client.workflows.wait_for_run(run.id)
print(run.step_results)   # counts, dwell, occupancy
How it works

Build it once, run it anywhere

Draw the pipeline in the editor, then run it on demand or wire it into your stack. No infrastructure to manage.

  1. Build the pipeline

    Drag together a source, a model, and the steps you need.

  2. Track and measure

    Add object tracking, counting, dwell time, and zone occupancy.

  3. Run on anything

    Point it at an image, a video, or a whole dataset.

  4. Deliver results

    Read results back, or push them to a signed webhook.

Built in

Tracking, counting, and analytics

The building blocks for turning detections into answers.

Object tracking Follow every object across frames with a stable ID.
Counting Count line crossings and entries over time.
Dwell and occupancy Measure how long objects stay and how full a zone gets.
Webhook delivery Signed, retried delivery straight into your systems.
FAQ

Workflows FAQ

What is a Pictograph workflow?

A workflow is a saved graph that runs a model over an image, a video, or a whole dataset: source, model, filter, track, step, visualize, sink. It loads ONNX weights per run, so it needs no standing deployment, and bills per frame processed.

How is workflow inference priced?

Workflows bill per frame processed at the underlying GPU rate plus the platform markup, charged from compute credits. You see a deposit up front and a true-up to actual usage when the run finishes, so you only pay for frames actually inferred.

Can workflows track objects across video frames?

Yes. Workflows include ByteTrack and BoT-SORT trackers, so a track ID follows each detection across frames. You can then add line-crossing counts, dwell time, and zone-occupancy steps on top of the tracks.

What can a workflow output?

A workflow can write annotated video, per-frame and per-track step results, and fire an outbound webhook on completion. Large result sets are offloaded to downloadable artifacts so nothing is silently truncated.

Build a vision pipeline today

$5/mo free compute. No payment method required.