deepstream-dev

What problem does it solve?

DeepStream 9.0 pipeline development with pyservicemaker becomes error-prone when users misconfigure sources, sinks, nvinfer settings, and DeepStream-specific metadata iteration rules. This Skill provides guardrails and correct configuration patterns so video analytics pipelines run reliably instead of failing silently or building broken engines.

Core Features & Use Cases

• End-to-end pipeline construction guidance: Establishes the correct DeepStream 9.0 pipeline flow (nvurisrcbin/nvstreammux/nvinfer/nvosdbin/sink) with platform-aware renderer selection.
• Configuration correctness for nvinfer and TensorRT: Ensures YAML syntax, correct sections, and handling dynamic ONNX input shapes via required infer-dims settings.
• Reliability and performance safety rules: Prevents common failures such as sink pad linking mistakes, iterator misuse for metadata, missing pyservicemaker inside venvs, and async/deadlock issues with tee/dynamic sources.
• Use cases: Building video analytics pipelines for detection, optional tracking, optional Kafka messaging, and common troubleshooting for DeepStream/pyservicemaker integration.

Quick Start

Build a minimal DeepStream 9.0 pyservicemaker pipeline for a local or RTSP source, run primary detection with nvinfer, visualize results with nvosdbin, and render using the correct platform-specific sink settings and pad linking syntax.