C# OpenCV 部署RecRecNet广角图像畸变矫正

天天代码码天天2024-06-18 14:33

C# OpenCV 部署RecRecNet广角图像畸变矫正

目录

说明

效果

模型信息

项目

代码

下载

说明

ICCV2023 - RecRecNet: Rectangling Rectified Wide-Angle Images by Thin-Plate Spline Model and DoF-based Curriculum Learning

参考：

https://github.com/KangLiao929/RecRecNet

https://github.com/hpc203/recrecnet-opencv-dnn

效果

模型信息

Model Properties

Inputs

name：input

tensor：Float[1, 3, 256, 256]

Outputs

name：output

tensor：Float[1, 162]

项目

代码

using OpenCvSharp;

using OpenCvSharp.Dnn;

using System;

using System.Collections.Generic;

using System.Drawing;

using System.Linq;

using System.Windows.Forms;

namespace OpenCvSharp_DNN_Demo

{

public partial class frmMain : Form

{

public frmMain()

{

InitializeComponent();

}

string fileFilter = "图片|*.bmp;*.jpg;*.jpeg;*.tiff;*.tiff;*.png";

string image_path = "";

DateTime dt1 = DateTime.Now;

DateTime dt2 = DateTime.Now;

int inpHeight;

int inpWidth;

string modelpath;

int grid_h = 8;

int grid_w = 8;

Mat grid;

Mat W_inv;

Net opencv_net;

Mat BN_image;

Mat image;

Mat result_image;

private void button1_Click(object sender, EventArgs e)

{

OpenFileDialog ofd = new OpenFileDialog();

ofd.Filter = fileFilter;

if (ofd.ShowDialog() != DialogResult.OK) return;

pictureBox1.Image = null;

pictureBox2.Image = null;

textBox1.Text = "";

image_path = ofd.FileName;

pictureBox1.Image = new Bitmap(image_path);

image = new Mat(image_path);

}

private void Form1_Load(object sender, EventArgs e)

{

modelpath = "model/model_deploy.onnx";

inpHeight = 256;

inpWidth = 256;

opencv_net = CvDnn.ReadNetFromOnnx(modelpath);

Common.get_norm_rigid_mesh_inv_grid(ref grid, ref W_inv, inpHeight, inpWidth, grid_h, grid_w);

image_path = "test_img/10.jpg";

pictureBox1.Image = new Bitmap(image_path);

}

private unsafe void button2_Click(object sender, EventArgs e)

{

if (image_path == "")

{

return;

}

textBox1.Text = "检测中，请稍等......";

pictureBox2.Image = null;

Application.DoEvents();

image = new Mat(image_path);

dt1 = DateTime.Now;

Mat img = new Mat();

Cv2.Resize(image, img, new OpenCvSharp.Size(inpWidth, inpHeight));

img.ConvertTo(img, MatType.CV_32FC3, 1.0f / 127.5f, -1.0f);

BN_image = CvDnn.BlobFromImage(img);

//配置图片输入数据

opencv_net.SetInput(BN_image);

//模型推理，读取推理结果

Mat[] outs = new Mat[1] { new Mat() };

string[] outBlobNames = opencv_net.GetUnconnectedOutLayersNames().ToArray();

opencv_net.Forward(outs, outBlobNames);

dt2 = DateTime.Now;

float* offset = (float*)outs[0].Data;

Mat tp = new Mat();

Mat ori_mesh_np_x = new Mat();

Mat ori_mesh_np_y = new Mat();

Common.get_ori_rigid_mesh_tp(tp, ori_mesh_np_x, ori_mesh_np_y, offset, inpHeight, inpWidth, grid_h, grid_w);

Mat T = W_inv * tp;

T = T.T();

Mat T_g = T * grid;

Mat output_tps = Common._interpolate(BN_image, T_g, new OpenCvSharp.Size(inpWidth, inpHeight));

Mat rectangling_np = (output_tps + 1) * 127.5;

rectangling_np.ConvertTo(rectangling_np, MatType.CV_8UC3);

Mat input_np = (img + 1) * 127.5;

List<Mat> outputs = new List<Mat>();

outputs.Add(rectangling_np);

outputs.Add(input_np);

outputs.Add(ori_mesh_np_x);

outputs.Add(ori_mesh_np_y);

Mat input_with_mesh = Common.draw_mesh_on_warp(outputs[1], outputs[2], outputs[3]);

Cv2.CvtColor(outputs[0], outputs[0], ColorConversionCodes.BGR2RGB);

Cv2.ImShow("mesh", input_with_mesh);

result_image = outputs[0].Clone();

pictureBox2.Image = new Bitmap(result_image.ToMemoryStream());

textBox1.Text = "推理耗时:" + (dt2 - dt1).TotalMilliseconds + "ms";

}

private void pictureBox2_DoubleClick(object sender, EventArgs e)

{

Common.ShowNormalImg(pictureBox2.Image);

}

private void pictureBox1_DoubleClick(object sender, EventArgs e)

{

Common.ShowNormalImg(pictureBox1.Image);

}

}

}

复制代码 
using OpenCvSharp;
using OpenCvSharp.Dnn;
using System;
using System.Collections.Generic;
using System.Drawing;
using System.Linq;
using System.Windows.Forms;

namespace OpenCvSharp_DNN_Demo
{
    public partial class frmMain : Form
    {
        public frmMain()
        {
            InitializeComponent();
        }

        string fileFilter = "图片|*.bmp;*.jpg;*.jpeg;*.tiff;*.tiff;*.png";
        string image_path = "";

        DateTime dt1 = DateTime.Now;
        DateTime dt2 = DateTime.Now;

        int inpHeight;
        int inpWidth;
        string modelpath;

        int grid_h = 8;
        int grid_w = 8;
        Mat grid;
        Mat W_inv;

        Net opencv_net;
        Mat BN_image;

        Mat image;
        Mat result_image;

        private void button1_Click(object sender, EventArgs e)
        {
            OpenFileDialog ofd = new OpenFileDialog();
            ofd.Filter = fileFilter;
            if (ofd.ShowDialog() != DialogResult.OK) return;

            pictureBox1.Image = null;
            pictureBox2.Image = null;
            textBox1.Text = "";

            image_path = ofd.FileName;
            pictureBox1.Image = new Bitmap(image_path);
            image = new Mat(image_path);
        }

        private void Form1_Load(object sender, EventArgs e)
        {
            modelpath = "model/model_deploy.onnx";

            inpHeight = 256;
            inpWidth = 256;

            opencv_net = CvDnn.ReadNetFromOnnx(modelpath);

            Common.get_norm_rigid_mesh_inv_grid(ref grid, ref W_inv, inpHeight, inpWidth, grid_h, grid_w);

            image_path = "test_img/10.jpg";
            pictureBox1.Image = new Bitmap(image_path);

        }

        private unsafe void button2_Click(object sender, EventArgs e)
        {
            if (image_path == "")
            {
                return;
            }
            textBox1.Text = "检测中，请稍等......";
            pictureBox2.Image = null;
            Application.DoEvents();

            image = new Mat(image_path);
            dt1 = DateTime.Now;

            Mat img = new Mat();

            Cv2.Resize(image, img, new OpenCvSharp.Size(inpWidth, inpHeight));

            img.ConvertTo(img, MatType.CV_32FC3, 1.0f / 127.5f, -1.0f);

            BN_image = CvDnn.BlobFromImage(img);

            //配置图片输入数据
            opencv_net.SetInput(BN_image);

            //模型推理，读取推理结果
            Mat[] outs = new Mat[1] { new Mat() };
            string[] outBlobNames = opencv_net.GetUnconnectedOutLayersNames().ToArray();

            opencv_net.Forward(outs, outBlobNames);

            dt2 = DateTime.Now;

            float* offset = (float*)outs[0].Data;

            Mat tp = new Mat();
            Mat ori_mesh_np_x = new Mat();
            Mat ori_mesh_np_y = new Mat();
            Common.get_ori_rigid_mesh_tp(tp, ori_mesh_np_x, ori_mesh_np_y, offset, inpHeight, inpWidth, grid_h, grid_w);
            Mat T = W_inv * tp;   
            T = T.T();    

            Mat T_g = T * grid;

            Mat output_tps = Common._interpolate(BN_image, T_g, new OpenCvSharp.Size(inpWidth, inpHeight));
            Mat rectangling_np = (output_tps + 1) * 127.5;
            rectangling_np.ConvertTo(rectangling_np, MatType.CV_8UC3);
            Mat input_np = (img + 1) * 127.5;

            List<Mat> outputs = new List<Mat>();
            outputs.Add(rectangling_np);
            outputs.Add(input_np);
            outputs.Add(ori_mesh_np_x);
            outputs.Add(ori_mesh_np_y);

            Mat input_with_mesh = Common.draw_mesh_on_warp(outputs[1], outputs[2], outputs[3]);

            Cv2.CvtColor(outputs[0], outputs[0], ColorConversionCodes.BGR2RGB);

            Cv2.ImShow("mesh", input_with_mesh);

            result_image = outputs[0].Clone();
            pictureBox2.Image = new Bitmap(result_image.ToMemoryStream());
            textBox1.Text = "推理耗时:" + (dt2 - dt1).TotalMilliseconds + "ms";

        }

        private void pictureBox2_DoubleClick(object sender, EventArgs e)
        {
            Common.ShowNormalImg(pictureBox2.Image);
        }

        private void pictureBox1_DoubleClick(object sender, EventArgs e)
        {
            Common.ShowNormalImg(pictureBox1.Image);
        }
    }
}

下载

源码下载

相关推荐
IT古董37 分钟前
第四章:大模型（LLM)】06.langchain原理-(3)LangChain Prompt 用法
java·人工智能·python
TGITCIC2 小时前
AI Search进化论：从RAG到DeepSearch的智能体演变全过程
人工智能·ai大模型·ai智能体·ai搜索·大模型ai·deepsearch·ai search
lucky_lyovo5 小时前
自然语言处理NLP---预训练模型与 BERT
人工智能·自然语言处理·bert
fantasy_arch5 小时前
pytorch例子计算两张图相似度
人工智能·pytorch·python
AndrewHZ7 小时前
【3D重建技术】如何基于遥感图像和DEM等数据进行城市级高精度三维重建？
图像处理·人工智能·深度学习·3d·dem·遥感图像·3d重建
飞哥数智坊7 小时前
Coze实战第18讲：Coze+计划任务，我终于实现了企微资讯简报的定时推送
人工智能·coze·trae
Code_流苏7 小时前
AI热点周报（8.10~8.16）：AI界“冰火两重天“，GPT-5陷入热议，DeepSeek R2模型训练受阻？
人工智能·gpt·gpt5·deepseek r2·ai热点·本周周报
赴3357 小时前
矿物分类案列 （一）六种方法对数据的填充
人工智能·python·机器学习·分类·数据挖掘·sklearn·矿物分类
大模型真好玩7 小时前
一文深度解析OpenAI近期发布系列大模型：意欲一统大模型江湖？
人工智能·python·mcp
双翌视觉8 小时前
工业视觉检测中的常见的四种打光方式
人工智能·计算机视觉·视觉检测
热门推荐
01UV安装并设置国内源02Qwen3-Coder 快速上手教程 | Qwen Code + Claude Code03【2025.08.06最新版】Android Studio下载、安装及配置记录（自动下载sdk）04KGG转MP3工具|非KGM文件|解密音频05蜘蛛磁力 搜索引擎大全，如何使用蜘蛛磁力查找磁力链接062025最新国内服务器可用docker源仓库地址大全（2025年8月更新）07NVIDIA显卡驱动、CUDA、cuDNN 和 TensorRT 版本匹配指南08TRAE 规则（Rules）配置指南：个人习惯、团队规范与最佳实践09阿里开源首个图像生成基础模型——Qwen-Image本地部署教程，超强中文渲染能力刷新SOTA！10TRAE Rules 实践：为项目配置 6A 工作流