ZED SDK 2.0

What’s New

The ZED SDK 2.0 release includes a major update of the SDK, Positional Tracking module and Unity plugin. It also introduces a new Spatial Mapping module and support for Jetson TX2.

2.0.0

New Features

Major SDK update

• The ZED SDK is now composed of different modules: Video, Depth, Positional Tracking and Spatial Mapping. Each module can be configured through dedicated parameters. Support functions have also been introduced and memory management has been simplified. The new API is much easier to use and integrate.
• ZED SDK 2.0 introduces breaking changes. To help you switch from 1.2 to 2.0, read the migration guide.

New Spatial Mapping module

• New Spatial Mapping module is introduced in 2.0.
• Spatial Mapping combines stereo depth sensing and motion tracking to capture a 3D map of the environment. Spatial mapping is useful for collision avoidance, motion planning and realistic blending of real and virtual worlds.

Major update of Positional Tracking module

• Refactored Positional Tracking module.
• Introduced new Spatial Memory mode that enables the ZED to memorize its surroundings. It significantly improves positional tracking by correcting drift and ensures a consistent experience across uses. Spatial Memory is enabled by default and replaces the previous Area Learning mode.

SDK

Updated Video module

• Added control of camera white balance.
• Reduced video capture latency on Linux: It is now possible to adjust the size of the buffer queue size for video capture using InitParams::reqBuffersCountLinux.
• Added real-time SVO mode: This mode reads SVO files and skips frames if necessary to simulate the behavior of the camera operating in live mode. Real-time mode is not available for SVO files recorded in RAW compression.
• Fixed bug with camera control functions that sometimes returned erroneous values.

Updated Depth module

• Improved depth quality for STANDARD mode and reduced aliasing.

Updated Positional Tracking module

• Introduced Spatial Memory mode that significantly reduces drift over time.
• Introduced new Pose, Translation, Orientation, Rotation and Transform classes that provide better and finer control over the API.
• Refactored tracking parameters into a TrackingParameters class. Use default parameters for optimal results.

Updated Spatial Mapping module

• Introduced new sl::Mesh class to store polygonal meshes.
• enableSpatialMapping() starts Spatial Mapping.
• requestMeshAsync(), getMeshRequestStatusAsync(), retrieveMeshAsync() request and retrieve the mesh that is being created. These functions are asynchronous and designed for real-time mesh extraction.
• extractWholeMesh() extracts the current mesh synchronously. This function is designed to extract the whole mesh at the end of a mapping session.
• disableSpatialMapping() stops Spatial Mapping.

Added new Mat class

• The sl::Mat class simplifies CPU and GPU memory management.
• Memory can be owned or shared (reference) with another Mat.
• Reallocation is handled by the SDK. It is not necessary to pre-allocate memory.
• Transfer between CPU and GPU memory is simplified using Mat::updateCPUfromGPU() and Mat::updateGPUfromCPU().
• Direct conversion to OpenCV cv::Mat is now available using Mat::toCVMat().
• Direct data access is available using Mat::getPtr().

General SDK changes

• Added default Camera constructor and open()/close() functions.
• Unified namespace: all the class and function are now into the single unified namespace “sl”.
• Changed filename input type from std::string to sl::String, based on const char* to allow Debug build under Windows.
• Renamed coordinate systems and added new COORDINATE_SYSTEM “LEFT_HANDED_Z_UP”.
• Fixed bug with invalid depth values when out of range.
• Minor bug fixes and general improvements.

Dependencies

• Removed Eigen.

Platform Support

• New Jetson TX2 is now supported.