savant framework
The launch of the new version of Savant 0.2.5 was announced, which is based on the new Nvidia DeepStream 6.3 SDK, as well as with improvements in adapters, encoders and more.
For those unfamiliar with the framework, you should know that this takes care of all the work with GStreamer or FFmpeg, allowing you to focus on creating optimized output pipelines using declarative (YAML) syntax and Python functions.
Learned hides all internals of Gstreamer from the developer and provides practical tools for implementing streaming AI applications in real life. Uses the standard Nvidia PeopleNet model to detect people and their faces and especially in places where privacy regulations apply, the framework allows tracking and blurring of faces.
What's new in Savant 0.2.5?
In this new version that is presented of Savant 0.2.5 as already mentioned The main novelty is the base in the new SDK de NVIDIA DeepStream 6.3 which includes bug fixes and improvements in video encoding.
And with the new version, the JPEG encoder now supports dGPU assisted encoding, since in previous versions, the JPEG encoder was software based. Fixed H.264 encoder issue.
On the part of the improvements to developer tools, It is highlighted that Savant now supports OpenTelemetry, which allows each frame processed by a pipeline to be instrumented, in addition to profiling the code, assigning records and attributes to traces, and analyzing errors and bottlenecks in the pipeline.
In Savant 0.2.5 it stands out a synchronous and asynchronous client SDK which can be used to develop unit tests, full-featured integration tests, and custom adapters.
In addition to this, it is also noted that Savant now has a useful script for ingesting videos or images into the pipeline directly from the development environment without the need to start separate adapters.
On the other hand, it stands out new adapter for data processing using Kafka/Redis, which enable the scalable processing needed to handle many streams in high-load, scalable environments. Adapters handle metadata with Kafka and video c
Also a multi-stream source test adapter, that allows measuring pipeline performance and tuning parameters, we implemented a utility adapter that ingests multiple flows into the pipeline. The adapter can create a desired parallel workload to help understand how the pipeline behaves under the workload and discover the optimal performance settings.
Of the other changes that stand out of this new version of Savant 0.2.5
- Python SDK to develop your own adapters.
- The RTSP video streaming adapter has the ability to process multiple streams simultaneously.
- HEVC encoding for industrial GigE cameras, with which streams recovered from GigE Vision cameras can be encoded with HEVC and sent with narrowband channels if necessary.
- Face re-identification using YOLOV8-Face, AdaFace and HNSWlib.
- Gender and age prediction based on YOLOV8-Face, MobileNet V2.
- Real-time instance segmentation (100+ FPS) based on YOLOV8M-Seg.
- 14 bug reports were closed.
A transition was made from a data transfer protocol based on Apache AVRO to a protocol based on Rust Rkyv (allowed to reduce GIL contention, increase the speed of serialization and deserialization). - Improved the way the Docker image
- Added support for element groups that allow you to combine data processing variations within a pipeline.
- Implemented advanced logging configuration with support for different levels for different components.
- Added H.264 software encoder for A100, V100, A40, Nvidia Jetson Orin Nano cards.
- Added a conveyor overload detector.
finally if you are interested in knowing more about it about this new version, you can check the details In the following link.