IMM

using System;
using System.IO;
using Tensorflow;
using Tensorflow.NumPy;
using static Tensorflow.Binding;
using Tensorflow.Keras.Engine;
using Tensorflow.Keras.Layers;
using Tensorflow.Keras.Models;
using Tensorflow.Keras.Utils;

class Program
{
static void Main(string[] args)
{
string modelPath = "ssd_mobilenet_v2.pb"; // 這裡放預訓練模型
string imagePath = "test.jpg"; // 測試圖片

// 讀取 TensorFlow 模型
var graph = new Graph().as_default();
var session = new Session(graph);

// 載入模型
byte[] modelBytes = File.ReadAllBytes(modelPath);
session.graph.Import(modelBytes);

// 讀取圖片並轉換為 Tensor
NDArray imageArray = LoadImage(imagePath);
Tensor inputTensor = graph.OperationByName("image_tensor").output(0);
Tensor detectionBoxes = graph.OperationByName("detection_boxes").output(0);
Tensor detectionScores = graph.OperationByName("detection_scores").output(0);
Tensor detectionClasses = graph.OperationByName("detection_classes").output(0);

// 執行物件偵測
var results = session.run(new[] { detectionBoxes, detectionScores, detectionClasses }, new FeedItem(inputTensor, imageArray));

// 處理結果
ProcessResults(results);
}

static NDArray LoadImage(string path)
{
byte[] imageBytes = File.ReadAllBytes(path);
NDArray imageArray = np.array(imageBytes);
imageArray = np.reshape(imageArray, new Shape(1, 300, 300, 3)); // 轉換為 TensorFlow 可接受格式
return imageArray;
}

static void ProcessResults(NDArray[] results)
{
Console.WriteLine("偵測到的物件:");
for (int i = 0; i < results[0].shape[1]; i++)
{
float score = results[1][0, i];
if (score > 0.5) // 只顯示高信心度的物件
{
Console.WriteLine($"物件 {i}: 信心度 {score * 100}%");
}
}
}
}