Skip to content

Open SVO2

The open_svo2 library provides tools for reading raw video and sensor data from SVO2 files, which is Stereolabs' "proprietary" MCAP-based container format for ZED camera recordings.

Add this dependency to your pyproject.toml:

pyproject.toml
dependencies = ["open_svo2 @ git+ssh://github.com/RadarML/open-svo2.git@main"]

Add as a submodule with git submodule add git@github.com:RadarML/open-svo2.git, then:

pyproject.toml
dependencies = ["open_svo2"]

[tool.uv.sources]
open_svo2 = { path = "open-svo2" }

Usage

In addition to the Python API, the svo2 command-line tool can be used to convert SVO2 files to standard formats. For example, to extract the raw H.265 video stream:

svo2 recording.svo2 output.h265

The output format is inferred from the file extension, or can be specified explicitly with --mode. Three modes are currently supported:

Mode Description
mp4 Extract the raw H.265 video stream and save it as an MP4 file.
h265 Extract the raw H.265 video stream and save it as a raw H.265 bitstream.
npz Extract the raw IMU data and save it as a NumPy archive.

Why use this tool?

The official Stereolabs SDK only provides decoded frames through retrieveImage(). This tool allows you to:

  • Extract raw H.265/HEVC bitstreams directly without decoding
  • Access video data without installing the ZED SDK
  • Work with the raw encoded data for custom processing pipelines

See Also