YOLOv5_Seg.md

YOLOv5-Seg usage

NOTE: The yaml file is not required.

• Convert model
• Edit the config_infer_primary_yoloV5_seg file

Convert model

1. Download the YOLOv5 repo and install the requirements

git clone https://github.com/ultralytics/yolov5.git
cd yolov5
pip3 install -r requirements.txt
pip3 install onnx onnxsim onnxruntime

NOTE: It is recommended to use Python virtualenv.

2. Copy conversor

Copy the export_yoloV5_seg.py file from DeepStream-Yolo-Seg/utils directory to the yolov5 folder.

3. Download the model

Download the pt file from YOLOv5 releases (example for YOLOv5s-Seg 7.0)

wget https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s-seg.pt

NOTE: You can use your custom model.

4. Convert model

Generate the ONNX model file (example for YOLOv5s-Seg)

python3 export_yoloV5_seg.py -w yolov5s-seg.pt --dynamic

NOTE: Confidence threshold (example for conf-thres = 0.25)

The minimum detection confidence threshold is configured in the ONNX exporter file. The pre-cluster-threshold should be >= the value used in the ONNX model.

--conf-thres 0.25

NOTE: NMS IoU threshold (example for iou-thres = 0.45)

--iou-thres 0.45

NOTE: Maximum detections (example for max-det = 100)

--max-det 100

NOTE: To convert a P6 model

--p6

NOTE: To change the inference size (defaut: 640 / 1280 for --p6 models)

-s SIZE
--size SIZE
-s HEIGHT WIDTH
--size HEIGHT WIDTH

Example for 1280

-s 1280

or

-s 1280 1280

NOTE: To simplify the ONNX model

--simplify

NOTE: To use dynamic batch-size (DeepStream >= 6.1)

--dynamic

NOTE: To use static batch-size (example for batch-size = 4)

--batch 4

5. Copy generated files

Copy the generated ONNX model file and labels.txt file (if generated) to the DeepStream-Yolo-Seg folder.

Edit the config_infer_primary_yoloV5_seg file

Edit the config_infer_primary_yoloV5_seg.txt file according to your model (example for YOLOv5s-Seg)

[property]
...
onnx-file=yolov5s-seg.onnx
model-engine-file=yolov5s-seg.onnx_b1_gpu0_fp32.engine
...

NOTE: To output the masks, use

[property]
...
output-instance-mask=1
segmentation-threshold=0.5
...

NOTE: The YOLOv5-Seg resizes the input with center padding. To get better accuracy, use

[property]
...
maintain-aspect-ratio=1
symmetric-padding=1
...