效果
项目
模型信息
Model Properties
Inputs
name：pixel_values
tensor：Float[-1, 3, -1, -1]
Outputs
name：reconstruction
tensor：Float[-1, 3, -1, -1]
代码
using Microsoft.ML.OnnxRuntime
using Microsoft.ML.OnnxRuntime.Tensors
using OpenCvSharp
using System
using System.Collections.Generic
using System.Drawing
using System.Drawing.Imaging
using System.Linq
using System.Windows.Forms
namespace Onnx_Demo
{
public partial class Form1 : Form
{
public Form1()
{
InitializeComponent()
}
string fileFilter = "*.*|*.bmp;*.jpg;*.jpeg;*.tiff;*.tiff;*.png"
string image_path = ""
string startupPath
DateTime dt1 = DateTime.Now
DateTime dt2 = DateTime.Now
string model_path
Mat image
Mat result_image
SessionOptions options
InferenceSession onnx_session
Tensor<float> input_tensor
List<NamedOnnxValue> input_container
IDisposableReadOnlyCollection<DisposableNamedOnnxValue> result_infer
DisposableNamedOnnxValue[] results_onnxvalue
Tensor<float> result_tensor
private void button1_Click(object sender, EventArgs e)
{
OpenFileDialog ofd = new OpenFileDialog()
ofd.Filter = fileFilter
if (ofd.ShowDialog() != DialogResult.OK) return
pictureBox1.Image = null
image_path = ofd.FileName
pictureBox1.Image = new Bitmap(image_path)
textBox1.Text = ""
image = new Mat(image_path)
pictureBox2.Image = null
}
private void button2_Click(object sender, EventArgs e)
{
if (image_path == "")
{
return
}
button2.Enabled = false
pictureBox2.Image = null
textBox1.Text = ""
Application.DoEvents()
// 读取原始图像（BGR）
image = new Mat(image_path)
int originalWidth = image.Cols
int originalHeight = image.Rows
// ------------------ 预处理 ------------------
// 1. 转换为RGB
Mat rgb = new Mat()
Cv2.CvtColor(image, rgb, ColorConversionCodes.BGR2RGB)
// 2. 调整尺寸使宽高均为4的倍数（APISR要求输入能被4整除）
int padHeight = (int)Math.Ceiling((double)rgb.Height / 4) * 4
int padWidth = (int)Math.Ceiling((double)rgb.Width / 4) * 4
Mat padded = new Mat()
Cv2.CopyMakeBorder(rgb, padded, 0, padHeight - rgb.Height, 0, padWidth - rgb.Width, BorderTypes.Constant, new Scalar(0, 0, 0))
// 3. 归一化到 [0,1] 并转换为浮点
padded.ConvertTo(padded, MatType.CV_32FC3, 1.0 / 255.0)
// 4. 构建 CHW 张量
int height = padded.Height
int width = padded.Width
Mat[] channels = Cv2.Split(padded)
List<float> dataList = new List<float>()
for (int c = 0
{
float[] channelData = new float[height * width]
System.Runtime.InteropServices.Marshal.Copy(channels[c].Data, channelData, 0, height * width)
dataList.AddRange(channelData)
}
float[] inputData = dataList.ToArray()
input_tensor = new DenseTensor<float>(inputData, new[] { 1, 3, height, width })
// 输入容器清空并添加
input_container.Clear()
input_container.Add(NamedOnnxValue.CreateFromTensor("pixel_values", input_tensor))
// ------------------ 推理 ------------------
dt1 = DateTime.Now
result_infer = onnx_session.Run(input_container)
dt2 = DateTime.Now
// 获取输出（修改为 float 类型）
results_onnxvalue = result_infer.ToArray()
result_tensor = results_onnxvalue[0].AsTensor<float>()
int[] outShape = result_tensor.Dimensions.ToArray()
int outChannels = outShape[1]
int outHeight = outShape[2]
int outWidth = outShape[3]
float[] predFloat = result_tensor.ToArray()
// 创建 OpenCV Mat 存储输出图像（RGB顺序）
Mat outputRgb = new Mat(outHeight, outWidth, MatType.CV_32FC3)
// 将 float 数组按 CHW 写入 Mat (OpenCV 是 HWC)
int index = 0
for (int h = 0
{
for (int w = 0
{
Vec3f pixel
pixel.Item0 = predFloat[index]
pixel.Item1 = predFloat[index + outHeight * outWidth]
pixel.Item2 = predFloat[index + 2 * outHeight * outWidth]
outputRgb.Set<Vec3f>(h, w, pixel)
index++
}
}
// 将像素值从 [0,1] 转换到 [0,255] 并转为 8UC3
outputRgb.ConvertTo(outputRgb, MatType.CV_8UC3, 255.0)
// 裁剪掉之前填充的部分（因为输出尺寸是填充后尺寸的4倍）
int cropHeight = originalHeight * 4
int cropWidth = originalWidth * 4
Rect roi = new Rect(0, 0, cropWidth, cropHeight)
Mat cropped = new Mat(outputRgb, roi)
// 将 RGB 转换为 BGR 以便 OpenCV 显示
Mat resultBgr = new Mat()
Cv2.CvtColor(cropped, resultBgr, ColorConversionCodes.RGB2BGR)
result_image = resultBgr.Clone()
// 显示结果
pictureBox2.Image = new Bitmap(result_image.ToMemoryStream())
textBox1.Text = $"推理耗时:{(dt2 - dt1).TotalMilliseconds:F2}ms 输出尺寸:{cropWidth}x{cropHeight}";
button2.Enabled = true;
}
private void Form1_Load(object sender, EventArgs e)
{
startupPath = System.Windows.Forms.Application.StartupPath;
model_path = "model/4x_APISR_GRL_GAN_generator.onnx";
// 创建会话，使用 CPU（可根据需要改为 CUDA）
options = new SessionOptions();
options.LogSeverityLevel = OrtLoggingLevel.ORT_LOGGING_LEVEL_INFO;
options.AppendExecutionProvider_CPU(0);
// 如需 GPU，取消下面注释并安装 Microsoft.ML.OnnxRuntime.Gpu
// options.AppendExecutionProvider_CUDA(0);
onnx_session = new InferenceSession(model_path, options);
input_container = new List<NamedOnnxValue>();
// 测试图片路径（可选）
image_path = "test_img/test01.png";
if (System.IO.File.Exists(image_path))
{
pictureBox1.Image = new Bitmap(image_path);
image = new Mat(image_path);
}
}
private void pictureBox1_DoubleClick(object sender, EventArgs e)
{
Common.ShowNormalImg(pictureBox1.Image);
}
private void pictureBox2_DoubleClick(object sender, EventArgs e)
{
Common.ShowNormalImg(pictureBox2.Image);
}
SaveFileDialog sdf = new SaveFileDialog();
private void button3_Click(object sender, EventArgs e)
{
if (result_image == null || result_image.Empty())
{
MessageBox.Show("请先进行推理！");
return;
}
sdf.Title = "保存超分图像";
sdf.Filter = "PNG图片 (*.png)|*.png|JPEG图片 (*.jpg)|*.jpg|BMP图片 (*.bmp)|*.bmp";
sdf.FilterIndex = 1;
if (sdf.ShowDialog() == DialogResult.OK)
{
string ext = System.IO.Path.GetExtension(sdf.FileName).ToLower();
ImageFormat format = ImageFormat.Png;
if (ext == ".jpg" || ext == ".jpeg")
format = ImageFormat.Jpeg;
elseif (ext == ".bmp")
format = ImageFormat.Bmp;
using (var stream = result_image.ToMemoryStream())
using (var bitmap = new Bitmap(stream))
{
bitmap.Save(sdf.FileName, format);
}
MessageBox.Show("保存成功，位置：" + sdf.FileName);
}
}
}
}
参考
github.com/Kiteretsu77…
