Skip to main content

Live Camera Demo (Beta)

Snapshot SDK for Android includes a beta live camera demo in the sample app. The demo provides Stream-like automatic license plate recognition (ALPR) from the device camera. It is intended for evaluation and development.

Open the Live Camera Demo​

After the Snapshot SDK is initialized, the sample app displays a menu with these options:

  • Live Camera: Opens the live camera demo.
  • Run Recognition (File): Runs recognition on a sample image set stored in the app assets.
  • Run Recognition (Bitmap Buffer): For each sample image, it loads it into a buffer and then runs recognition.

Select Live Camera and grant camera access when prompted. The demo uses the device's rear camera.

What the Demo Shows​

Live camera screen showing a bounding box around a detected license plate, confidence score, and a list of confirmed plates with make, model, and color

The live camera screen displays:

  • A live camera preview.
  • A bounding box, recognized plate, and confidence score for current detections.
  • A list of recent confirmed plates, including their best plate crop, confidence, region, number of sightings, make, model, color, and detection time.
  • The effective recognition frame rate (FPS).
  • An MMC toggle. When your license includes the MMC feature, enabling it adds available vehicle make, model, and color information to confirmed results.

How It Works​

The demo processes camera frames locally with Snapshot SDK. It uses a latest-frame backpressure model: while Snapshot is processing a frame, newly arriving frames replace older unprocessed frames instead of being queued. This keeps the preview responsive and avoids building a backlog, but the displayed FPS is the effective processed-frame rate, not the camera sensor frame rate.

Frame rate depends on device performance, image conditions, and enabled recognition features. In internal testing, the camera could provide up to 12 frames per second, while Snapshot processing could reduce the effective rate to approximately 4 frames per second.

The demo stabilizes repeated detections before adding a plate to the confirmed-results list. This helps reduce transient or duplicate results from consecutive frames.

Lookup Consumption​

Beta lookup usage

This feature uses a Snapshot SDK lookup for every processed frame. It behaves similarly to a stream, but it is not Plate Recognizer Stream and does not use Stream's lookup model.

Depending on Android device performance, the demo typically consumes approximately 4 to 12 lookups per second. Evaluate expected lookup consumption before using this feature beyond testing or development.

Configure Recognition​

The sample can be configured with region hints and Snapshot SDK engine parameters. For supported regions, parameters, and response fields, see the Snapshot.engineProcBuffer().

For Android-specific initialization and image-processing guidance, see the Integration Guide.