本文へ移動
cccskills
無料GitHub で公開

deepstream-yolo

Routes DeepStream-Yolo deployment, model conversion, multi-GIE, and INT8 workflows for supported YOLO-family models.

VectorSpaceLab/AREX-Skill
インストール方法を見る

含まれるファイル（115）

  • SKILL.md4.6 KB
  • assets/configs/config_infer_primary_codetr.txt709 B
  • assets/configs/config_infer_primary_damoyolo.txt655 B
  • assets/configs/config_infer_primary_dfine.txt643 B
  • assets/configs/config_infer_primary_goldyolo.txt689 B
  • assets/configs/config_infer_primary_ppyoloe_plus.txt681 B
  • assets/configs/config_infer_primary_ppyoloe.txt707 B
  • assets/configs/config_infer_primary_rfdetr.txt648 B
  • assets/configs/config_infer_primary_rtdetr.txt663 B
  • assets/configs/config_infer_primary_rtmdet.txt765 B
  • assets/configs/config_infer_primary_yolo11.txt681 B
  • assets/configs/config_infer_primary_yolo26.txt658 B
  • assets/configs/config_infer_primary_yolomaster.txt692 B
  • assets/configs/config_infer_primary_yolonas_custom.txt669 B
  • assets/configs/config_infer_primary_yolonas.txt689 B
  • assets/configs/config_infer_primary_yolor.txt683 B
  • assets/configs/config_infer_primary_yoloV10.txt659 B
  • assets/configs/config_infer_primary_yoloV12.txt682 B
  • assets/configs/config_infer_primary_yoloV13.txt682 B
  • assets/configs/config_infer_primary_yoloV2.txt724 B
  • assets/configs/config_infer_primary_yoloV5.txt681 B
  • assets/configs/config_infer_primary_yoloV5u.txt682 B
  • assets/configs/config_infer_primary_yoloV6.txt681 B
  • assets/configs/config_infer_primary_yoloV7.txt680 B
  • assets/configs/config_infer_primary_yoloV8.txt681 B
  • assets/configs/config_infer_primary_yoloV9.txt682 B
  • assets/configs/config_infer_primary_yolox_legacy.txt711 B
  • assets/configs/config_infer_primary_yolox.txt661 B
  • assets/configs/config_infer_primary.txt744 B
  • assets/configs/deepstream_app_config.txt863 B
  • assets/configs/labels_rfdetr.txt692 B
  • assets/configs/labels.txt624 B
  • assets/images/multipleGIEs_tree.png24.4 KB
  • assets/nvdsinfer_custom_impl_Yolo/calibrator.cpp4.1 KB
  • assets/nvdsinfer_custom_impl_Yolo/calibrator.h2.2 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/activation_layer.cpp5.0 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/activation_layer.h362 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/batchnorm_layer.cpp2.5 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/batchnorm_layer.h489 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/channels_layer.cpp1.2 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/channels_layer.h397 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/convolutional_layer.cpp5.1 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/convolutional_layer.h549 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/deconvolutional_layer.cpp5.1 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/deconvolutional_layer.h555 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/implicit_layer.cpp1.2 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/implicit_layer.h443 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/pooling_layer.cpp1.9 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/pooling_layer.h359 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/reorg_layer.cpp4.3 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/reorg_layer.h379 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/route_layer.cpp2.6 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/route_layer.h388 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/sam_layer.cpp795 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/sam_layer.h413 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/shortcut_layer.cpp1.2 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/shortcut_layer.h504 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/slice_layer.cpp2.7 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/slice_layer.h380 B
  • assets/nvdsinfer_custom_impl_Yolo/layers/upsample_layer.cpp1.0 KB
  • assets/nvdsinfer_custom_impl_Yolo/layers/upsample_layer.h362 B
  • assets/nvdsinfer_custom_impl_Yolo/Makefile3.0 KB
  • assets/nvdsinfer_custom_impl_Yolo/nvdsinfer_yolo_engine.cpp5.2 KB
  • assets/nvdsinfer_custom_impl_Yolo/nvdsparsebbox_Yolo_cuda.cu4.2 KB
  • assets/nvdsinfer_custom_impl_Yolo/nvdsparsebbox_Yolo.cpp4.3 KB
  • assets/nvdsinfer_custom_impl_Yolo/utils.cpp4.1 KB
  • assets/nvdsinfer_custom_impl_Yolo/utils.h1.8 KB
  • assets/nvdsinfer_custom_impl_Yolo/yolo.cpp24.1 KB
  • assets/nvdsinfer_custom_impl_Yolo/yolo.h4.4 KB
  • assets/nvdsinfer_custom_impl_Yolo/yoloForward_nc.cu3.5 KB
  • assets/nvdsinfer_custom_impl_Yolo/yoloForward_v2.cu4.3 KB
  • assets/nvdsinfer_custom_impl_Yolo/yoloForward.cu3.6 KB
  • assets/nvdsinfer_custom_impl_Yolo/yoloPlugins.cpp9.6 KB
  • assets/nvdsinfer_custom_impl_Yolo/yoloPlugins.h5.9 KB
  • references/installation.md1.9 KB
  • references/model-family-matrix.md4.7 KB
  • references/repo-provenance.md1.2 KB
  • references/repo-routing-metadata.json347 B
  • references/runtime-assets.md1.3 KB
  • references/troubleshooting.md2.9 KB
  • scripts/build-nvdsinfer-plugin.sh3.0 KB
  • scripts/check-deepstream-toolchain.sh1.0 KB
  • scripts/make-calibration-list.sh765 B
  • scripts/stage-runtime-tree.sh1.8 KB
  • sub-skills/deployment/references/configuration.md2.6 KB
  • sub-skills/deployment/references/troubleshooting.md1.3 KB
  • sub-skills/deployment/references/workflows.md2.6 KB
  • sub-skills/deployment/scripts/build-nvdsinfer-plugin.sh3.0 KB
  • sub-skills/deployment/scripts/check-deepstream-toolchain.sh1.0 KB
  • sub-skills/deployment/SKILL.md3.4 KB
  • sub-skills/int8-benchmarking/references/benchmark-notes.md1.1 KB
  • sub-skills/int8-benchmarking/references/troubleshooting.md1.2 KB
  • sub-skills/int8-benchmarking/references/workflows.md1.7 KB
  • sub-skills/int8-benchmarking/scripts/build-nvdsinfer-plugin.sh3.0 KB
  • sub-skills/int8-benchmarking/scripts/make-calibration-list.sh765 B
  • sub-skills/int8-benchmarking/SKILL.md2.7 KB
  • sub-skills/model-conversion/references/model-family-matrix.md3.2 KB
  • sub-skills/model-conversion/references/troubleshooting.md1.4 KB
  • sub-skills/model-conversion/references/upstream-dependencies.md1.6 KB
  • sub-skills/model-conversion/references/workflows.md2.0 KB
  • sub-skills/model-conversion/scripts/export_rtdetr_ultralytics.py4.0 KB
  • sub-skills/model-conversion/scripts/export_yolo11.py4.2 KB
  • sub-skills/model-conversion/scripts/export_yolo26.py4.8 KB
  • sub-skills/model-conversion/scripts/export_yolomaster.py4.2 KB
  • sub-skills/model-conversion/scripts/export_yoloV10.py4.8 KB
  • sub-skills/model-conversion/scripts/export_yolov12.py4.2 KB
  • sub-skills/model-conversion/scripts/export_yoloV13.py4.2 KB
  • sub-skills/model-conversion/scripts/export_yoloV5u.py4.2 KB
  • sub-skills/model-conversion/scripts/export_yoloV8.py4.2 KB
  • sub-skills/model-conversion/SKILL.md3.4 KB
  • sub-skills/multi-gie/references/configuration.md1.8 KB
  • sub-skills/multi-gie/references/troubleshooting.md1.2 KB
  • sub-skills/multi-gie/references/workflows.md1.9 KB
  • sub-skills/multi-gie/scripts/setup-multi-gie-tree.sh6.3 KB
  • sub-skills/multi-gie/SKILL.md2.4 KB

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

DeepStream-Yolo

Use this skill for DeepStream YOLO deployment work: build the custom parser library, choose the right config template, convert weights to ONNX, wire multiple GIEs, and prepare INT8 calibration runs.

Start here

  1. Read references/repo-provenance.md if you need to know whether this skill still matches the current checkout.
  2. Run scripts/check-deepstream-toolchain.sh before any build or runtime attempt.
  3. If you want to inspect or edit the bundled runtime tree first, run scripts/stage-runtime-tree.sh --output-dir ./deepstream-yolo-runtime.
  4. Read references/model-family-matrix.md before changing a config template.
  5. Use the sub-skill that matches the task family:
    • sub-skills/deployment/SKILL.md for single-model build/run work.
    • sub-skills/model-conversion/SKILL.md for exporter selection and ONNX conversion.
    • sub-skills/multi-gie/SKILL.md for multiple primary/secondary detectors.
    • sub-skills/int8-benchmarking/SKILL.md for calibration and benchmark tuning.

What this skill covers

  • DeepStream app configuration for one detector.
  • Custom parser library builds in nvdsinfer_custom_impl_Yolo.
  • Model-family-specific config templates and their special knobs.
  • Ultralytics-family exporter scripts and labels generation.
  • Multi-GIE folder scaffolding and plugin-version updates.
  • INT8 calibration image lists, OpenCV build switch, and benchmark notes.

Shared references

  • references/installation.md — read when you need the host prerequisites, CUDA_VER mapping, or a quick reminder that DeepStream runtime is separate from the Python inspection environment.
  • references/runtime-assets.md — read when you need to know where the bundled parser source, configs, and multi-GIE layout image live.
  • scripts/stage-runtime-tree.sh — stage the bundled parser source and configs into a fresh runtime tree before running or editing them.
  • scripts/build-nvdsinfer-plugin.sh — stage the bundled runtime tree and build the custom parser library on a DeepStream host.
  • references/model-family-matrix.md — read when you need to match a model family to a config template or understand the important special knobs.
  • references/troubleshooting.md — read when build, runtime, export, or calibration errors need a cross-cutting fix.
  • references/repo-routing-metadata.json — used by the repo-skill router during import and discovery.
  • references/repo-provenance.md — read when checking staleness or refreshing this skill.

Runtime expectations

  • The generated skill is self-contained; do not depend on the original source checkout being preserved.
  • DeepStream runtime commands need an NVIDIA DeepStream host or container; this generated skill can still document the workflow even when that SDK is absent on the current machine.
  • The bundled assets/ tree contains the parser source, build files, config templates, and multi-GIE illustration needed for self-contained runtime staging.
  • The Ultralytics exporter path was inspected in a temporary Python environment with torch, ultralytics, onnx, onnxslim, and onnxruntime installed.

Route summary

deployment

Use this for:

  • building nvdsinfer_custom_impl_Yolo
  • choosing and editing a single-model config_infer_primary*.txt
  • wiring deepstream_app_config.txt for one detector
  • troubleshooting DeepStream build and runtime issues
  • understanding how custom models and labels fit into the pipeline

model-conversion

Use this for:

  • converting Ultralytics-family .pt checkpoints to ONNX
  • deciding which exporter script to use
  • generating labels.txt
  • understanding which upstream repos are still reference-only in this skill

multi-gie

Use this for:

  • duplicating the repo layout for multiple detectors
  • changing YOLOLAYER_PLUGIN_VERSION
  • wiring secondary-gie entries and operate-on-gie-id
  • moving engine files into the correct gieN/ folder

int8-benchmarking

Use this for:

  • enabling INT8 calibration support with OPENCV=1
  • preparing calibration.txt and INT8_CALIB_IMG_PATH
  • reading the benchmark table and tuning NMS values
  • deciding when calibration is blocked by missing OpenCV dev headers

Quick reminders

  • Set CUDA_VER before building.
  • Use the config template that matches the model family, not just the checkpoint name.
  • Keep labels.txt synchronized with the exporter output or the class list.
  • Do not treat CPU-only inspection as proof that DeepStream runtime works.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

ログインしてレビューを書く

同じリポジトリのスキル

概要と使いどころ

3d-resnets-pytorch

無料

Routes 3D ResNets PyTorch video action-recognition workflows across training, inference, and data preparation.

日本語の概要は準備中です。原文の説明を表示しています。

VectorSpaceLab/AREX-Skill3322026年9月3日 更新

3ddfa

無料

Guide 3DDFA Python inference, geometry rendering, training/evaluation, and optional C++ ONNX workflows for 3D dense face alignment.

日本語の概要は準備中です。原文の説明を表示しています。

VectorSpaceLab/AREX-Skill3322026年9月3日 更新

3ddfa-v2

無料

Routes 3DDFA_V2 face-alignment setup, still-image demos, video tracking, and ONNX benchmarking workflows.

日本語の概要は準備中です。原文の説明を表示しています。

VectorSpaceLab/AREX-Skill3322026年9月3日 更新

ab3dmot

無料

Operate AB3DMOT 3D multi-object tracking workflows for KITTI and nuScenes data, tracking, evaluation, and visualization.

日本語の概要は準備中です。原文の説明を表示しています。

VectorSpaceLab/AREX-Skill3322026年9月3日 更新

accelerate

無料

Use Hugging Face Accelerate for PyTorch training-loop migration, distributed launch/configuration, DeepSpeed/FSDP/TPU backend setup, big-model inference/offload, checkpointing, tracking, and troubleshooting.

日本語の概要は準備中です。原文の説明を表示しています。

VectorSpaceLab/AREX-Skill3322026年9月3日 更新

acme

無料

Route Acme reinforcement-learning framework tasks across core loops, replay/data, JAX agents, and TensorFlow/Sonnet agents.

日本語の概要は準備中です。原文の説明を表示しています。

VectorSpaceLab/AREX-Skill3322026年9月3日 更新

VectorSpaceLab のスキルをすべて見る

このスキルの問題を報告する