Model Conversion

1. YOLO (Benchmark)

Benchmark testing refers to the Ultralytics team's benchmark testing of YOLO on ten different model formats, measuring speed and accuracy. These formats include: PyTorch, TorchScript, ONNX, OpenVINO, TF SavedModel, TF GraphDef, TF Lite, MNN, NCNN, ExecuTorch. Since the official only provided comparison charts for yolo11 on JetsonOrin Super, we use yolo11's test results as reference here. It can be seen that TensorRT format models have the fastest inference speed.

Comparison Chart

2. Model Conversion

Use yolo's CLI command in the terminal for conversion

yolo export model=yolo26n.pt format=engine # creates 'yolo26n.engine'

This command will convert models into two different formats: .onnx and .engine, where .onnx is a universal format

Besides yolo26n.pt, the four models for segmentation, pose detection, classification, and OBB can also be converted to .engine format using the same method.

xxxxxxxxxx
yolo export model=yolo26n-seg.pt format=engine
yolo export model=yolo26n-pose.pt format=engine
yolo export model=yolo26n-obb.pt format=engine
yolo export model=yolo26n-cls.pt format=engine