Module data-replay-camera

Replay images from Viam Data Management as a camera.

Model viam-modules:camera:data-replay

A camera component that returns images stored in Viam Data Management. Instead of streaming live sensor data, it replays previously captured images by returning the next matching image on every get_image() call.

---

When to use this module

Use data-replay-camera when you want a camera-like component that:

• Feeds a dataset (or filtered slice of Data Management) into a pipeline that expects a camera
• Replays images deterministically (good for debugging, demos, and repeatable tests)

This module is not intended to be a high-throughput streaming camera.

---

Quick start

Configuration

Attribute template

{
  "dataset_id": "<string>",
  "tags": ["<string>"],
  "labels": ["<string>"]
}

Example configuration

{
  "dataset_id": "demo",
  "tags": ["demo"],
  "labels": []
}

Attributes

Name	Type	Inclusion	Description
dataset_id	string	Optional	Dataset ID to filter images.
tags	list	Optional	Tag filter for images.
labels	list	Optional	Bounding-box label filter for images.

Authentication: This module automatically uses Viam-provided credentials when running as a module. No API keys need to be configured manually.

---

How to use it

Camera API

Implements the RDK Camera API:

• get_image(): returns the next image (wraps around after the last)
• get_images(): returns a single image per call (internally calls get_image())

The camera returns images filtered by the configured dataset_id, tags, and labels. If multiple filters are provided, results must match all constraints.

Example usage:

camera.get_image()

---

Behavior

• Replay order: Each call returns the next image from the matched result set.
• Wrap-around: After the last image is returned, the next call returns the first image again.
• Caching: The module caches the list of matching images the first time a given filter combination is requested.
• Refreshing results: If dataset contents / tags / labels change, restart or reconfigure the component to refresh cached results.
• Encoding: Images are returned as JPEG.
• No matches: If the filter returns no images, get_image() raises an error.

---

Prerequisites / links

• Tag images and create a dataset (tutorial):
  • https://docs.viam.com/data-ai/train/create-dataset/
• Data Management service overview:
  • https://docs.viam.com/data-ai/capture-data/capture-sync/
• Dataset label data:
  • https://docs.viam.com/data-ai/train/annotate-images/