ONNX Runtime in Computer Vision

Introduction

ONNX Runtime helps you run machine learning models fast and easily on many devices. It makes using models from different tools simple.

You want to run a trained model on your computer without retraining.

You have a model from another tool and want to use it in your app.

You want faster predictions on images or text using a ready model.

You want to run the same model on different devices like PC or phone.

You want to test how well a model works on new data quickly.

Syntax

Computer Vision

import onnxruntime as ort

# Load the model
session = ort.InferenceSession('model.onnx')

# Prepare input data as a dictionary
inputs = {'input_name': input_array}

# Run the model to get outputs
outputs = session.run(None, inputs)

Replace 'model.onnx' with your ONNX model file path.

Input names must match the model's expected input names.

Examples

Load a handwritten digit model and run it on random image data.

Computer Vision

import onnxruntime as ort
import numpy as np

session = ort.InferenceSession('mnist.onnx')
input_name = session.get_inputs()[0].name
input_data = np.random.rand(1, 1, 28, 28).astype('float32')
outputs = session.run(None, {input_name: input_data})

Run the model and print the first output array.

Computer Vision

import onnxruntime as ort

session = ort.InferenceSession('model.onnx')
input_name = session.get_inputs()[0].name
inputs = {input_name: your_numpy_array}
outputs = session.run(None, inputs)
print(outputs[0])

Sample Model

This code loads an ONNX model that adds 1 to each input number. It runs the model on three numbers and prints the results.

Computer Vision

import onnxruntime as ort
import numpy as np

# Load a simple ONNX model (for example, a model that adds 1 to input)
# Here we create a dummy input and run the model
session = ort.InferenceSession('add_one.onnx')

# Get the input name from the model
input_name = session.get_inputs()[0].name

# Create input data: a batch of 3 numbers
input_data = np.array([[1.0], [2.0], [3.0]], dtype=np.float32)

# Run the model
outputs = session.run(None, {input_name: input_data})

# Print the output
print('Input:', input_data)
print('Output:', outputs[0])

Important Notes

ONNX Runtime supports many hardware accelerations for faster results.

Always check the input and output names with session.get_inputs() and session.get_outputs().

Input data must match the model's expected shape and data type.

Summary

ONNX Runtime lets you run machine learning models easily on many devices.

You load a model, prepare inputs, and get outputs with simple Python code.

It works well for quick testing and deploying models without extra training.