marcoslucianops / DeepStream-Yolo

NVIDIA DeepStream SDK 6.3 / 6.2 / 6.1.1 / 6.1 / 6.0.1 / 6.0 / 5.1 implementation for YOLO models

DeepStream-Yolo

NVIDIA DeepStream SDK 6.3 / 6.2 / 6.1.1 / 6.1 / 6.0.1 / 6.0 / 5.1 configuration for YOLO models

---

YOLO-Pose: https://github.com/marcoslucianops/DeepStream-Yolo-Pose

YOLO-Seg: https://github.com/marcoslucianops/DeepStream-Yolo-Seg

YOLO-Face: https://github.com/marcoslucianops/DeepStream-Yolo-Face

---

Important: please export the ONNX model with the new export file, generate the TensorRT engine again with the updated files, and use the new config_infer_primary file according to your model

---

Future updates

• DeepStream tutorials
• Updated INT8 calibration
• Support for classification models

Improvements on this repository

• Support for INT8 calibration
• Support for non square models
• Models benchmarks
• Support for Darknet models (YOLOv4, etc) using cfg and weights conversion with GPU post-processing
• Support for RT-DETR, YOLO-NAS, PPYOLOE+, PPYOLOE, DAMO-YOLO, YOLOX, YOLOR, YOLOv8, YOLOv7, YOLOv6 and YOLOv5 using ONNX conversion with GPU post-processing
• GPU bbox parser (it is slightly slower than CPU bbox parser on V100 GPU tests)
• Support for DeepStream 5.1
• Custom ONNX model parser (NvDsInferYoloCudaEngineGet)
• Dynamic batch-size for Darknet and ONNX exported models
• INT8 calibration (PTQ) for Darknet and ONNX exported models
• New output structure (fix wrong output on DeepStream < 6.2) - it need to export the ONNX model with the new export file, generate the TensorRT engine again with the updated files, and use the new config_infer_primary file according to your model
• RT-DETR PyTorch (https://github.com/lyuwenyu/RT-DETR/tree/main/rtdetr_pytorch)
• RT-DETR Paddle (https://github.com/lyuwenyu/RT-DETR/tree/main/rtdetr_paddle)
• RT-DETR Ultralytics (https://docs.ultralytics.com/models/rtdetr)

Getting started

• Requirements
• Supported models
• Benchmarks
• dGPU installation
• Basic usage
• Docker usage
• NMS configuration
• INT8 calibration
• YOLOv5 usage
• YOLOv6 usage
• YOLOv7 usage
• YOLOv8 usage
• YOLOR usage
• YOLOX usage
• DAMO-YOLO usage
• PP-YOLOE / PP-YOLOE+ usage
• YOLO-NAS usage
• RT-DETR PyTorch usage
• RT-DETR Paddle usage
• RT-DETR Ultralytics usage
• Using your custom model
• Multiple YOLO GIEs

Requirements

DeepStream 6.3 on x86 platform

• Ubuntu 20.04
• CUDA 12.1 Update 1
• TensorRT 8.5 GA Update 2 (8.5.3.1)
• NVIDIA Driver 525.125.06 (Data center / Tesla series) / 530.41.03 (TITAN, GeForce RTX / GTX series and RTX / Quadro series)
• NVIDIA DeepStream SDK 6.3
• GStreamer 1.16.3
• DeepStream-Yolo

DeepStream 6.2 on x86 platform

• Ubuntu 20.04
• CUDA 11.8
• TensorRT 8.5 GA Update 1 (8.5.2.2)
• NVIDIA Driver 525.85.12 (Data center / Tesla series) / 525.105.17 (TITAN, GeForce RTX / GTX series and RTX / Quadro series)
• NVIDIA DeepStream SDK 6.2
• GStreamer 1.16.3
• DeepStream-Yolo

DeepStream 6.1.1 on x86 platform

• Ubuntu 20.04
• CUDA 11.7 Update 1
• TensorRT 8.4 GA (8.4.1.5)
• NVIDIA Driver 515.65.01
• NVIDIA DeepStream SDK 6.1.1
• GStreamer 1.16.2
• DeepStream-Yolo

DeepStream 6.1 on x86 platform

• Ubuntu 20.04
• CUDA 11.6 Update 1
• TensorRT 8.2 GA Update 4 (8.2.5.1)
• NVIDIA Driver 510.47.03
• NVIDIA DeepStream SDK 6.1
• GStreamer 1.16.2
• DeepStream-Yolo

DeepStream 6.0.1 / 6.0 on x86 platform

• Ubuntu 18.04
• CUDA 11.4 Update 1
• TensorRT 8.0 GA (8.0.1)
• NVIDIA Driver 470.63.01
• NVIDIA DeepStream SDK 6.0.1 / 6.0
• GStreamer 1.14.5
• DeepStream-Yolo

DeepStream 5.1 on x86 platform

• Ubuntu 18.04
• CUDA 11.1
• TensorRT 7.2.2
• NVIDIA Driver 460.32.03
• NVIDIA DeepStream SDK 5.1
• GStreamer 1.14.5
• DeepStream-Yolo

DeepStream 6.3 on Jetson platform

• JetPack 5.1.2
• NVIDIA DeepStream SDK 6.3
• DeepStream-Yolo

DeepStream 6.2 on Jetson platform

• JetPack 5.1.2 / 5.1.1 / 5.1
• NVIDIA DeepStream SDK 6.2
• DeepStream-Yolo

DeepStream 6.1.1 on Jetson platform

• JetPack 5.0.2
• NVIDIA DeepStream SDK 6.1.1
• DeepStream-Yolo

DeepStream 6.1 on Jetson platform

• JetPack 5.0.1 DP
• NVIDIA DeepStream SDK 6.1
• DeepStream-Yolo

DeepStream 6.0.1 / 6.0 on Jetson platform

• JetPack 4.6.4
• NVIDIA DeepStream SDK 6.0.1 / 6.0
• DeepStream-Yolo

DeepStream 5.1 on Jetson platform

• JetPack 4.5.1
• NVIDIA DeepStream SDK 5.1
• DeepStream-Yolo

Supported models

• Darknet
• MobileNet-YOLO
• YOLO-Fastest
• YOLOv5
• YOLOv6
• YOLOv7
• YOLOv8
• YOLOR
• YOLOX
• DAMO-YOLO
• PP-YOLOE / PP-YOLOE+
• YOLO-NAS

Basic usage

1. Download the repo

git clone https://github.com/marcoslucianops/DeepStream-Yolo.git
cd DeepStream-Yolo

2. Download the cfg and weights files from Darknet repo to the DeepStream-Yolo folder

3. Compile the lib

• DeepStream 6.3 on x86 platform

  CUDA_VER=12.1 make -C nvdsinfer_custom_impl_Yolo
  

• DeepStream 6.2 on x86 platform

  CUDA_VER=11.8 make -C nvdsinfer_custom_impl_Yolo
  

• DeepStream 6.1.1 on x86 platform

  CUDA_VER=11.7 make -C nvdsinfer_custom_impl_Yolo
  

• DeepStream 6.1 on x86 platform

  CUDA_VER=11.6 make -C nvdsinfer_custom_impl_Yolo
  

• DeepStream 6.0.1 / 6.0 on x86 platform

  CUDA_VER=11.4 make -C nvdsinfer_custom_impl_Yolo
  

• DeepStream 5.1 on x86 platform

  CUDA_VER=11.1 make -C nvdsinfer_custom_impl_Yolo
  

• DeepStream 6.3 / 6.2 / 6.1.1 / 6.1 on Jetson platform

  CUDA_VER=11.4 make -C nvdsinfer_custom_impl_Yolo
  

• DeepStream 6.0.1 / 6.0 / 5.1 on Jetson platform

  CUDA_VER=10.2 make -C nvdsinfer_custom_impl_Yolo
  

4. Edit the config_infer_primary.txt file according to your model (example for YOLOv4)

[property]
...
custom-network-config=yolov4.cfg
model-file=yolov4.weights
...

NOTE: For Darknet models, by default, the dynamic batch-size is set. To use static batch-size, uncomment the line

...
force-implicit-batch-dim=1
...

5. Run

deepstream-app -c deepstream_app_config.txt

NOTE: The TensorRT engine file may take a very long time to generate (sometimes more than 10 minutes).

NOTE: If you want to use YOLOv2 or YOLOv2-Tiny models, change the deepstream_app_config.txt file before run it

...
[primary-gie]
...
config-file=config_infer_primary_yoloV2.txt
...

Docker usage

• x86 platform

  nvcr.io/nvidia/deepstream:6.3-gc-triton-devel
  nvcr.io/nvidia/deepstream:6.3-triton-multiarch
  

• Jetson platform

  nvcr.io/nvidia/deepstream-l4t:6.3-samples
  nvcr.io/nvidia/deepstream:6.3-triton-multiarch
  

  NOTE: To compile the nvdsinfer_custom_impl_Yolo, you need to install the g++ inside the container

  apt-get install build-essential
  

  NOTE: With DeepStream 6.3, the docker containers do not package libraries necessary for certain multimedia operations like audio data parsing, CPU decode, and CPU encode. This change could affect processing certain video streams/files like mp4 that include audio track. Please run the below script inside the docker images to install additional packages that might be necessary to use all of the DeepStreamSDK features:

  /opt/nvidia/deepstream/deepstream/user_additional_install.sh
  

NMS Configuration

To change the nms-iou-threshold, pre-cluster-threshold and topk values, modify the config_infer file

[class-attrs-all]
nms-iou-threshold=0.45
pre-cluster-threshold=0.25
topk=300

NOTE: Make sure to set cluster-mode=2 in the config_infer file.

Extract metadata

You can get metadata from DeepStream using Python and C/C++. For C/C++, you can edit the deepstream-app or deepstream-test codes. For Python, your can install and edit deepstream_python_apps.

Basically, you need manipulate the NvDsObjectMeta (Python / C/C++) and NvDsFrameMeta (Python / C/C++) to get the label, position, etc. of bboxes.

My projects: https://www.youtube.com/MarcosLucianoTV