Attributions

RTEAM.ai's computer vision features are built on the open machine learning work listed below. We are grateful to its authors. Each entry preserves the title, author, source, and license that the respective license requires, and the list names every model the analytics pipeline loads and no others.

PeopleNet

• Used for: Person detection (headcount, entries and exits)
• Author: NVIDIA Corporation
• Source: https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/peoplenet
• License: NVIDIA Open Model License
• Note: Used as published (deployable_quantized_onnx_v2.6.3, INT8), compiled to a TensorRT engine for inference. Only the person class is used.

Tracking and video

Per-camera tracking uses NvDCF, the discriminative correlation filter tracker included in the NVIDIA DeepStream SDK, with appearance-based re-identification disabled and motion-only re-association enabled. No tracker model weights are used. Video decode, inference and the surrounding pipeline run on the NVIDIA DeepStream SDK, CUDA and TensorRT, under NVIDIA's licenses, and on GStreamer and GLib (LGPL-2.1, dynamically linked and unmodified).

Complete notices

The list above is the complete set of models that the analytics pipeline loads. The full third-party notices — including the license texts for the runtime libraries, the Eclipse Public/Distribution License election for the MQTT client, and the SIL Open Font License covering the Geist typeface used on this page — ship with the software as THIRD_PARTY_NOTICES.md and are available on request.