We’re all excited to announce the latest release of BrainFrame v0.28.1!
If you’ve already installed BrainFrame, simply run:
Then download an updated version of the client here.
If you haven’t installed BrainFrame, then follow our installation instructions.
- The BrainFrame client now has support for limiting the number of streams that are viewed. The client will now only show 5 streams, and will load streams dynamically when the user clicks on one. This allows the client to run on underpowered computers, even if the server is running 100+ streams.
Added support for
iGPUon OpenVINO Capsules (disabled by default, will be enabled by default in
v0.29release). Read below for more information on how to enable it.
- Upgraded the capsule system to use Tensorflow 2! (Specifically, 2.5.0). All capsules are forwards compatible.
- Improved API data throughput for the
- Upgraded BrainFrame docker images from Cuda 10 to Cuda 11. No changes in the host system should be necessary.
- GStreamer improvements:
- Added support for OpenCL
- Reduce memory usage when using NVCODEC
- Reduced memory usage of
journalmicroservice by ~3x
- Fix a memory leak that occurs when deleting streams
- Fix freezes that happen on the
/api/streams/statusesAPI which could cause timeout errors on the client
This feature is currently in beta. Here’s how to use it!
# 1. Stop the BrainFrame server brainframe compose down # 2. Add an environment variable telling BrainFrame to allow iGPU echo OPENVINO_DEVICE_PRIORITY=GPU,CPU >> $(brainframe info install_path)/.env # 3. Start BrainFrame brainframe compose up -d # 4. View logs, and look for the log "Loading capsule ___ onto devices GPU brainframe compose logs -f core
Make sure you have an OpenVINO capsule installed, to test this feature.