Table for neural network regression

tbl = readtable('yourFile.txt');
tbl.fileNames = cellstr(tbl.fileNames); % Ensure the first column is a cell array

Check the Table: Verify that the table has the correct structure:

disp(tbl);
% It should look like:
% fileNames                               Percent
% {'C:\path\to\image1.webp'}                68.6
% {'C:\path\to\image2.webp'}                68.6

Using the Table with trainNetwork

The first column (fileNames) should contain image paths, and the second column (Percent) should contain numeric regression targets. Pass the table to trainNetwork as follows:

regNet = trainNetwork(tbl, 'Percent', layers, options);

Here:

• 'Percent' specifies the column containing the regression target.
• The first column, fileNames, is implicitly recognized as the input data.

Common Issues and Fixes

1. File Paths Not Found: Ensure that the file paths in fileNames are correct and accessible from your system. Use exist(path, 'file') to verify each path:

assert(all(cellfun(@(x) exist(x, 'file') == 2, tbl.fileNames)), 'Some file paths are invalid.');

1. Image Formats: Verify that all images have a consistent format (e.g., .webp, .jpg) and are readable by MATLAB.

2. Table Column Order: The predictors must be in the first column, and the response variable must be in the specified target column (in this case, 'Percent').

3. Table Data Types: The fileNames column must be a cell array of strings, and the Percent column must be numeric.

Example

Here’s an example of creating and using a table for neural network regression:

% Example data
fileNames = {'C:\path\to\image1.webp'; 'C:\path\to\image2.webp'};
percent = [68.6; 72.1];

% Create table
tbl = table(fileNames, percent, 'VariableNames', {'fileNames', 'Percent'});

% Define a simple network
layers = [
    imageInputLayer([32 32 3])
    fullyConnectedLayer(1)
    regressionLayer
];
options = trainingOptions('sgdm', 'MaxEpochs', 5, 'Plots', 'training-progress');

% Train network
regNet = trainNetwork(tbl, 'Percent', layers, options);

This will train the network using the table data. Ensure your table format matches this structure.