关于使用 C#（Xamarin 开发环境）从浮点矩阵转换为 Java.Lang.Object 的问题

Question

Is it possible to convert a multidimensional float matrix to a Java.Lang.Object in C#?
The neural network I use (.tflite format) is correctly loaded and to make inference two Java.Lang.Object are requested as Input parameters, hereafter InputObj01 and InputObj02:

1. InputObj01 --> Input to be classified. In my case Matrix Dimensions: [1, 4000, 2]
2. InputObj02 --> Output Model. In my case Matrix Dimension: [1, 4000]

I found no problems converting InputObj02 to Java.Lang.Object, using the following steps:

float[][] Input02 = new float[][]{ new float[4000] };  
var InputObj02 = Arrays.FromArray<float[]>(Input02);

But if I try to use the same technique to get InputObj01, in the following way:

float[][,] Input01 = new float[][,]{ new float[4000, 2] };
var InputObj01 = Arrays.FromArray<float[,]>(Input01)

the following exception is thrown:

**System.InvalidCastException:** 'Specified cast is not valid.'

Using directive:

using System;
using Android.App;
using Android.Content.PM;
using Android.Runtime;
using Android.OS;
using Android.Widget;                                            
using System.IO;                                                
using Android.Content.Res;                                      
using Java.IO;                                                  
using Java.Nio.Channels;                                        
using Java.Nio;                                                 
using Xamarin.TensorFlow.Lite;
using Java.Util;

Answer 1

For your needs, it may be easier to just use a simple loop:

        float[][,] Input01 = new float[][,] { new float[4000, 2] };
        List<float[,]> newList = new List<float[,]>();

        foreach (float[,] array in Input01)
        {
                newList.Add(array);
        }