mzku/yolo11s-coco-onnx-dynamic-simplify-nms

Argus YOLO11s COCO ONNX Dynamic Simplify NMS

This repository contains a YOLO11s object detection model exported to ONNX for use with Argus, a C++ live object detection pipeline using FFmpeg, TensorRT, and hardware-accelerated video processing.

This is not an official Ultralytics repository.

Files

• yolo11s_dynamic_simplify_nms.onnx: ONNX model with dynamic shapes, graph simplification, and NMS included.

Expected Output

The model includes NMS and returns detections in the format:

[x1, y1, x2, y2, confidence, class_id]

Usage With Argus

Argus can build a TensorRT engine from this ONNX model and run batched object detection on live video streams.

Project repository: https://github.com/zraisan/argus

Export Command

This model can be recreated with Ultralytics using:

yolo export
model=yolo11s.pt
format=onnx
imgsz=640
opset=17
dynamic=True
simplify=True
nms=True
max_det=300
batch=16

License

This model is based on Ultralytics YOLO and is released under the AGPL-3.0 license.