skills.rest
DEEPX-AI
DEEPX-AIOFFICIAL
Explore all skills by this author →
WebsiteGitHub

dx-agent-stream-build-pipeline

Generates GStreamer pipeline applications for DEEPX NPU video inference with dx_stream.

3|5|Updated Jan 21, 2025
Search Tags:#pipeline#object detection#npu#gstreamer#video inference#dx_stream#deepx
One-click install
npx skills add https://github.com/DEEPX-AI/dx_stream --skill dx-agent-stream-build-pipeline-deepx-ai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: dx-agent-stream-build-pipeline
Source: https://github.com/DEEPX-AI/dx_stream/tree/main/.deepx/skills/dx-agent-stream-build-pipeline
Command: npx skills add https://github.com/DEEPX-AI/dx_stream --skill dx-agent-stream-build-pipeline-deepx-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Building GStreamer pipelines for DEEPX NPU inference requires correctly wiring many elements (dxpreprocess, dxinfer, dxpostprocess, dxtracker, dxosd) with matching IDs, absolute model paths, and postprocess libraries. This Skill generates complete, tested pipeline applications so you avoid common wiring and configuration errors. ## Core Features & Use Cases - Five Pipeline Categories: Generates single-model, multi-model, cascaded (secondary inference), parallel, and broker pipelines with complete templates for pipeline.py, run scripts, session.json, and README. - Model & Postprocess Mapping: Ships a reference table of 14 supported DXNN models (YOLO variants, SCRFD, EfficientNet) with their input sizes and matching postprocess shared libraries. - Isolated Output Sessions: Writes all generated applications into timestamped dx-agent-dev/ session directories, keeping production code untouched. - Use Case: Ask for a cascaded pipeline using yolo26n for detection and EfficientNet_Lite0 for classification, and receive a runnable session directory with a Python pipeline, shell wrapper, and documentation. ## Quick Start Ask the agent to build a dx_stream object detection pipeline using the yolo26n model with a sample video input.

Frequently Asked Questions about dx-agent-stream-build-pipeline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a GStreamer pipeline for DEEPX NPU inference?▼

Use this Skill to generate a complete pipeline application by specifying the pipeline category, DXNN model name, and input source. It produces a pipeline.py script, a run shell wrapper, session.json metadata, and a README inside an isolated dx-agent-dev session directory.

What models does dx_stream support for object detection?▼

The Skill supports 14 DXNN models including yolo26n, YoloV5S, YoloV7, YoloV8N, YoloV9S, YoloXS, and YOLOV11N for detection, plus pose, segmentation, face detection, and classification models. Each model maps to a specific postprocess shared library.

How do I create a cascaded pipeline with secondary inference?▼

Request the cascaded category, which adds a secondary inference stage with secondary-mode=true on dxpreprocess, dxinfer, and dxpostprocess. The primary detector output branches through a tee, secondary results merge via dxgather, and session.json records both models and postprocess libraries.

Why does my dx_stream pipeline fail with pydxs import errors?▼

The pydxs module is only available inside the dx_stream virtual environment. Activate it with source venv-dx_stream/bin/activate from the dx_stream root before running pipeline.py, or reinstall with pip install -e . if the venv exists but pydxs is missing.

Can I run dx_stream pipelines headless without a display?▼

Yes, generated pipeline.py scripts always include a --headless flag that switches the sink to fakesink, and they also check the DISPLAY environment variable as a fallback. File recording via x264enc with tune=zerolatency is supported through the --output argument.

Related Skills

Explore More Software Engineering Skills→
easingthemeseasingthemes
6

dx-eject

Copy plugin assets into the consumer project and archive originals.

Community
Advanced
sandeepmvlsandeepmvl
21

redis-streams-rails

Implement Redis Streams event pipelines with consumer groups for Rails applications.

Community
Advanced
PP-NamiasPP-Namias
2

sanity-schema-validator

Validate Sanity schemas against best practices and detect common issues.

Community
Intermediate
NVIDIANVIDIA
3.2k

deepstream-generate-pipeline

Generate validated gst-launch-1.0 pipelines for NVIDIA DeepStream video inference and tracking.

Official
Advanced