[ Tech stack ]
NVIDIA DeepStream
NVIDIA's SDK for analyzing dozens of video streams in real time on a GPU.
DeepStream assembles video analytics pipelines as GStreamer graphs where every stage stays on the GPU: NVDEC hardware decoding, multi-stream batching, TensorRT or Triton inference, object tracking, on-screen display and metadata publishing to Kafka, MQTT or AMQP. The same pipeline, written in C++ or Python, runs on a Jetson at the edge or on a data-center GPU, with primary and secondary inference to detect and then classify. It's our choice when several cameras have to be analyzed continuously on site, without needless CPU copies.
[ Why NVIDIA DeepStream at Dexon ]
What this technology does well, and why we use it.
Typical usage: People and vehicle counting, industrial site security, in-line quality control, retail analytics.
- 01
Zero-copy GStreamer pipeline: NVDEC decoding, inference, tracking and encoding on the GPU.
- 02
nvstreammux multi-stream batching: dozens of RTSP cameras on a single card.
- 03
Built-in trackers (NvDCF, NvSORT, NvDeepSORT) and cascaded detect-then-classify inference.
- 04
Metadata published to Kafka, MQTT, AMQP or Azure IoT through nvmsgbroker.
[ Complementary technologies ]
The building blocks we often mobilise alongside.
A stack rarely exists alone. Here are the technologies Dexon most often pairs with this one, through pipeline habits, usage similarity or internal mastery. Click on a building block to see its scope.
[ Reassurance ]
- 0+
- custom projects delivered
- 30
- engineers, designers, project managers
- 80 %
- from top French schools
- 24 h
- average reply time
[ They trust us ]
More than 100 French and European companies trust us
[ Press ]
They talk about us.
Nationwide coverage by BFM Business, Le Figaro, Challenges, La Tribune and CNews. An outside reading of our work and our innovations.
