Skip to content

Releases: levipereira/triton-server-yolo

YOLOv9,YOLOv8,YOLOv7 Model Release - ONNX

03 Mar 21:56
4e19c36
Compare
Choose a tag to compare

YOLOv7 and YOLOv9 Series Model Release - ONNX

Description:

This release contains YOLOv7 and YOLOv9 models exported from the original reparametrized models.

All models were exported in their original size format, 640x640

Models with Dynamic Batching and TRT Efficient NMS plugin

Models with "end2end" in their name were exported with dynamic batching and include the TRT Efficient NMS plugin.

Evaluation Models

YOLOv9 / YOLOv9 QAT

  • eval-yolov9-c-end2end.onnx
  • eval-yolov9-e-end2end.onnx
  • eval-yolov9-c-qat-end2end.onnx
  • eval-yolov9-e-qat-end2end.onnx

Parameters Used:

python3 export.py \
 --weights <yolo>.pt \
 --imgsz 640 \
--dynamic \
--inplace \
--simplify \
--topk-all 300 \
--iou-thres 0.7 \
--conf-thres 0.001 \
--include onnx_end2end 

YOLOv7 / YOLOv7 QAT

  • eval-yolov7-end2end.onnx
  • eval-yolov7x-end2end.onnx
  • eval-yolov7-qat-end2end.onnx
  • eval-yolov7x-qat-end2end.onnx

Parameters Used:

python3 export.py \
--weights <model>.pt \
--img-size 640 \
--dynamic-batch \
--simplify  \
--grid \
--end2end \
--topk-all 300 \
--iou-thres 0.7 \
--conf-thres 0.001 

Inference Models

YOLOv9 / YOLOv9 QAT

  • infer-yolov9-c-end2end.onnx
  • infer-yolov9-e-end2end.onnx
  • infer-yolov9-c-qat-end2end.onnx
  • infer-yolov9-e-qat-end2end.onnx
    Parameters Used:
python3 export.py \
 --weights <yolo>.pt \
 --imgsz 640 \
--dynamic \
--inplace \
--simplify \
--topk-all 100 \
--iou-thres 0.45 \
--conf-thres 0.25 \
--include onnx_end2end 

YOLOv7 / YOLOv7 QAT

  • infer-yolov7-end2end.onnx
  • infer-yolov7x-end2end.onnx
  • infer-yolov7-qat-end2end.onnx
  • infer-yolov7x-qat-end2end.onnx
    Parameters Used:
python3 export.py \
--weights <model>.pt \
--img-size 640 \
--dynamic-batch \
--simplify  \
--grid \
--end2end \
--topk-all 100 \
--iou-thres 0.45 \
--conf-thres 0.25 

Models with Dynamic Batching

Models without "end2end" in their name were exported only with dynamic batching:

Inference Models YOLOv9 and YOLOv7

  • infer-yolov7.onnx
  • infer-yolov7x.onnx
  • infer-yolov9-c.onnx
  • infer-yolov9-e.onnx
  • infer-yolov9-c-qat.onnx
  • infer-yolov9-e-qat.onnx