Savant Explained: GigE Vision Camera Support

Savant has supported GigE Vision industrial cameras since version 0.2.0. Those cams are widely used in industrial environments when high-definition imaging is required allowing customers to fine-tune the parameters and calibrate optics very precisely.

Consider visiting our GitHub, and don’t forget to subscribe to our Twitter to receive updates on Savant. Also, we have Discord, where we help users to onboard.

GigE Vision support in Savant is a game-changer for industrial AI: the feature enables using Nvidia-accelerated computations on the industrial edge and in the core. Without Savant, industrial developers don’t have a straightforward solution for ingesting video streams from industrial cams to DeepStream. Even Jetson NX can handle a FullHD stream from a GigE Vision cam in real-time (30–40+ FPS) with Savant.

As for now, we only have customers who use Basler cameras, so we cannot be sure that all other GigE vision cams are supported. Our GigE adapter is implemented on the Aravis open-source toolkit.

Before the upcoming Savant 0.2.5, the GigE Vision Source adapter supported only raw exchange format unsuitable for building distributed systems communicating over the network. We recommend deploying a container accessing the cam and the AI module processing the stream on the same host if you need to handle raw images from a GigE Vision cam. The best way to implement their communication is to use the Unix domain sockets scheme, which Savant supports.

In the upcoming 0.2.5, the adapter supports CPU-based HEVC compression, which makes it possible to transfer the resulting video stream over the network. Remember that HEVC is either loss or lossless compression, depending on encoder capabilities. Savant uses x265enc, available in GStreamer, for video compression. Refer to the GStreamer documentation for details.

There is a sample demonstrating GigE Source adapter functionality with compression and without it in the Savant repository.