1.27.0

• New API: Vio::addVelocity(...) (in core C++ SDK)
• New documentation: https://spectacularai.github.io/docs/sdk/
• Orbbec Astra 2 support
• Orbbec Windows support
• Bug fixes

Known issues:

• Orbbec examples, vio_visu.py and vio_jsonl.cpp leave the camera in bad state when closing, causing tracking to fail when used second time.

1.26.2

• Support for Orbbec Femto Mega & Femto Bolt (BETA)
• A new addition to Mapping API, you can get sparse feature points used by SLAM via spectacularAI::mapping::Frame::sparseFeatures
• Fixes sai-cli process failing with Android data
• Fixes useVioAutoExposure = True

1.26.1

• Orbbec Astra 2 support (BETA)!
• OAK-D: Fixes crashes related to missing getTimestampDevice() method
• Improved sai-cli processing and preview visualizations
• Bugfixes

Known Issues:

• useVioAutoExposure = True is buggy, this issue also affects earlier releases to some degree

1.25.1

• Fix monocular video replay on some datasets

1.25.0

• New Python sai-cli command line tool that makes using the Spetacular AI tools and recording data more straight forward
• Better optimized ffmpeg settings for recording
• Bug fixes

1.24.1

• Fix issue preventing using PNG images as depth video on Windows

1.24.0

Azure Kinect plugin is here!

• Improved SLAM performance
• Enable RGB mapping API output from mono data with passthroughColorImages
• Fix segfault in Mapping API special case: mono mode with depth data

1.23.1

• Added RealSense recording tool & example

1.23.0

Highlights:

• Optimized settings for Gaussian Splatting and NeRF reconstructions
• Better warnings when incoming data has issues, such as too sparse IMU samples or arriving too late compared to camera data
• Improved tracking performance

Known issues:

• Extended output callback doesn't handle color images properly

1.22.1

• Exposes hooks API in the Depth AI plugin to Python, previously only accessible through C++