gusucode.com > vision工具箱matlab源码程序 > vision/+vision/+internal/+ocr/+tool/Session.m
classdef Session < handle
properties
FileName
IsChanged
CanTrain
IsTrained
IsInitialized
ImageSet
OCRLanguage
OCRCharacterSet
OutputDirectory
OutputLanguage
AutoLabel
Font
end
methods
function this = Session()
this.ImageSet = vision.internal.ocr.tool.ImageSet();
this.reset();
end
function reset(this)
this.FileName = '';
this.IsChanged = false;
this.CanTrain = false;
this.IsTrained = false;
this.IsInitialized = false;
this.AutoLabel = true;
this.OCRLanguage = 'English';
this.OCRCharacterSet = '';
this.OutputDirectory = '';
this.OutputLanguage = 'myLanguage';
this.ImageSet.reset();
end
%------------------------------------------------------------------
function updateFont(this, val)
this.Font = val;
% update ImageSet font too
this.ImageSet.Font = this.Font;
end
%------------------------------------------------------------------
function ret = hasAnyItems(this)
if this.ImageSet.CharView
ret = this.ImageSet.hasAnyCharacters();
else
ret = this.ImageSet.hasAnyImages();
end
end
%------------------------------------------------------------------
function ret = hasAnyUnknowns(this)
chars = this.ImageSet.CharMap.keys;
if any(strcmp(char(0), chars))
ret = true;
else
ret = false;
end
end
%------------------------------------------------------------------
function ret = hasAllUnknowns(this)
chars = this.ImageSet.CharMap.keys;
if all(strcmp(char(0), chars))
ret = true;
else
ret = false;
end
end
%------------------------------------------------------------------
function ret = hasAnyImages(this)
ret = this.ImageSet.hasAnyImages();
end
%------------------------------------------------------------------
function checkImagePaths(this, pathname, filename)
this.ImageSet.checkImagePaths(pathname, filename);
end
%------------------------------------------------------------------
function [imgSet, bboxData] = getTrainingData(this)
imgStruct = this.ImageSet.ImageStruct;
imgfiles = {imgStruct(:).imageFilename};
imgSet = vision.internal.ocr.tool.CustomImageSet(imgfiles);
imgSet.TextDetectionParams = this.ImageSet.TextDetectionParams;
% Description is used as font name.
imgSet.Description = this.OutputLanguage;
for i = 1:numel(imgStruct)
validBoxes = this.ImageSet.getValidBoxes(i);
validText = this.ImageSet.getValidText(i);
bboxData(i).bboxes = validBoxes;
bboxData(i).chars = validText;
bboxData(i).page = repelem(0,numel(bboxData(i).chars));
% Convert from spatial coordinates to pixel coordinates.
% In the app using spatial coordinates is simplier for
% visualizing box patches. Pixel coordinates are only
% required when it is time to train.
bboxData(i).bboxes(:,1:2) = bboxData(i).bboxes(:,1:2) + 0.5;
end
end
%------------------------------------------------------------------
function generateEvaluationFunction(this)
codeGenerator = vision.internal.calibration.tool.MCodeGenerator;
codeGenerator.addLine(sprintf(['%% Auto-generated by ', ...
'%s app on %s'], 'ocrTrainer', date));
codeGenerator.addLine(['%----------------------------------------', ...
'---------------']);
codeGenerator.addLine('% ');
codeGenerator.addLine('% evaluateOCRTraining Evaluate OCR training');
codeGenerator.addLine('% results = evaluateOCRTraining(I) runs OCR using the trained language model');
codeGenerator.addLine('% and returns an image, ocrI, that has been annotated with the recognition');
codeGenerator.addLine('% results.');
codeGenerator.addLine('% ');
codeGenerator.addLine('% ocrI = evaluateOCRTraining(..., roi) optionally limit OCR to a');
codeGenerator.addLine('% rectangular region of interest, roi, within I. Use this if I contains a');
codeGenerator.addLine('% lot of non-text background, which can hinder OCR performance.');
codeGenerator.addLine('% ');
codeGenerator.addLine('% [..., results] = evaluateOCRTraining(I) optionally returns the detailed');
codeGenerator.addLine('% recogition results in an ocrText object.');
codeGenerator.addLine('% ');
codeGenerator.addLine('% Notes');
codeGenerator.addLine('% -----');
codeGenerator.addLine('% ');
codeGenerator.addLine('% 1) You may need to pre-process your test images to remove noise and');
codeGenerator.addLine('% improve text segmentation.');
codeGenerator.addLine('% ');
codeGenerator.addLine('% 2) You may need to modify the OCR ''TextLayout'' parameter used below to');
codeGenerator.addLine('% something that is more suitable for your input images.');
codeGenerator.addLine('% ');
codeGenerator.addLine('% 3) You may need to re-train your language using more training samples.');
codeGenerator.addLine('% ');
codeGenerator.addLine('% 4) You may need to modify this function and pass a region of interest');
codeGenerator.addLine('% (ROI) to the OCR function. This is required if your image has a lot');
codeGenerator.addLine('% of non-text background. Alternatively, use IMCROP to manually crop the');
codeGenerator.addLine('% image before evaluating OCR.');
codeGenerator.addLine('% ');
codeGenerator.addLine('% Example 1 - Draw an ROI and evaluate OCR training');
codeGenerator.addLine('% --------------------------------------------------- ');
codeGenerator.addLine(sprintf('%% I = imread(''%s'');', this.ImageSet.ImageStruct(1).imageFilename));
codeGenerator.addLine('% figure');
codeGenerator.addLine('% imshow(I)');
codeGenerator.addLine('% ');
codeGenerator.addLine('% % Draw a region of interest');
codeGenerator.addLine('% h = imrect');
codeGenerator.addLine('% ');
codeGenerator.addLine('% % Evaluate OCR within ROI');
codeGenerator.addLine('% roi = h.getPosition;');
codeGenerator.addLine('% ocrI = evaluateOCRTraining(I, roi);');
codeGenerator.addLine('% ');
codeGenerator.addLine('% % Show results');
codeGenerator.addLine('% figure');
codeGenerator.addLine('% imshow(ocrI)');
codeGenerator.addLine('% ');
codeGenerator.addLine('% Example 2 - Get all OCR results');
codeGenerator.addLine('% -------------------------------');
codeGenerator.addLine(sprintf('%% I = imread(''%s'');', this.ImageSet.ImageStruct(1).imageFilename));
codeGenerator.addLine('% [ocrI, results] = evaluateOCRTraining(I);');
codeGenerator.addLine('% results.Text');
codeGenerator.addLine('% results.CharacterConfidences');
codeGenerator.addLine('% ');
codeGenerator.addLine('% Example 3 - Batch OCR evaluation ');
codeGenerator.addLine('% --------------------------------');
codeGenerator.addLine('% % Save this generated function and use it in the imageBatchProcessor');
codeGenerator.addLine('% imageBatchProcessor');
codeGenerator.addLine('% ');
codeGenerator.addLine('% See also ocr, ocrTrainer, ocrText');
codeGenerator.addLine('function [ocrI, results] = evaluateOCRTraining(I, roi)');
lang = fullfile(this.OutputDirectory,...,
this.OutputLanguage , 'tessdata', ...
[this.OutputLanguage '.traineddata']);
codeGenerator.addComment('Location of trained OCR language data');
codeGenerator.addLine(sprintf('trainedLanguage = ''%s'';', lang));
codeGenerator.addLine(' ');
codeGenerator.addLine('% Run OCR using trained language. You may need to modify OCR parameters or');
codeGenerator.addLine('% pre-process your test images for optimal results. Also, consider');
codeGenerator.addLine('% specifying an ROI input to OCR in case your images have a lot of non-text');
codeGenerator.addLine('% background.');
codeGenerator.addLine('layout = ''Block'';');
codeGenerator.addLine('if nargin == 2');
codeGenerator.addLine(' results = ocr(I, roi, ...');
codeGenerator.addLine(' ''Language'', trainedLanguage, ...');
codeGenerator.addLine(' ''TextLayout'', layout); ');
codeGenerator.addLine('else');
codeGenerator.addLine(' results = ocr(I, ...');
codeGenerator.addLine(' ''Language'', trainedLanguage, ...');
codeGenerator.addLine(' ''TextLayout'', layout); ');
codeGenerator.addLine('end');
codeGenerator.addLine(' ');
codeGenerator.addLine('ocrI = insertOCRAnnotation(I, results);');
codeGenerator.addLine(' ');
codeGenerator.addLine('%--------------------------------------------------------------------------');
codeGenerator.addLine('% Annotate I with OCR results.');
codeGenerator.addLine('%--------------------------------------------------------------------------');
codeGenerator.addLine('function J = insertOCRAnnotation(I, results)');
codeGenerator.addLine('text = results.Text;');
codeGenerator.addLine(' ');
codeGenerator.addLine('I = im2uint8(I);');
codeGenerator.addLine('if isempty(deblank(text))');
codeGenerator.addLine(' % Text not recognized.');
codeGenerator.addLine(' text = ''Unable to recognize any text.'';');
codeGenerator.addLine(' [M,N,~] = size(I);');
codeGenerator.addLine(' J = insertText(I, [N/2 M/2], text, ...');
codeGenerator.addLine(sprintf(' ''AnchorPoint'', ''Center'', ''FontSize'', 24, ''Font'', ''%s'');', this.Font));
codeGenerator.addLine(' ');
codeGenerator.addLine('else');
codeGenerator.addLine(' location = results.CharacterBoundingBoxes;');
codeGenerator.addLine(' ');
codeGenerator.addLine(' % Remove new lines from results.');
codeGenerator.addLine(' newlines = text == char(10);');
codeGenerator.addLine(' text(newlines) = [];');
codeGenerator.addLine(' location(newlines, :) = [];');
codeGenerator.addLine(' ');
codeGenerator.addLine(' % Remove spaces from results');
codeGenerator.addLine(' spaces = isspace(text);');
codeGenerator.addLine(' text(spaces) = [];');
codeGenerator.addLine(' location(spaces, :) = [];');
codeGenerator.addLine(' ');
codeGenerator.addLine(' % Convert text array into cell array of strings.');
codeGenerator.addLine(' text = num2cell(text);');
codeGenerator.addLine(' ');
codeGenerator.addLine(' % Pad the image to help annotate results close to the image border.');
codeGenerator.addLine(' I = padarray(I, [50 50], uint8(255));');
codeGenerator.addLine(' location(:,1:2) = location(:,1:2) + 50;');
codeGenerator.addLine(' ');
codeGenerator.addLine(' % Insert text annotations.');
codeGenerator.addLine(' J = insertObjectAnnotation(I, ''rectangle'', location, text);');
codeGenerator.addLine('end');
content = codeGenerator.CodeString;
% Open template code in an untitled file in the editor
editorDoc = matlab.desktop.editor.newDocument(content);
editorDoc.smartIndentContents;
editorDoc.goToLine(1);
end
end
methods(Hidden)
%------------------------------------------------------------------
% Utility function to generated tesseract box files.
%------------------------------------------------------------------
function writeBoxFiles(this)
[trainingImages, trainingData] = this.getTrainingData();
fontnames = {trainingImages(:).Description};
language = this.OutputLanguage;
for i = 1:numel(trainingImages) % array of imageSet, 1 per font
for j = 1:trainingImages(i).Count
I = read(trainingImages(i),j);
baseFilename = sprintf('%s.%s.exp%d',language, fontnames{i},j);
% write box file given training boxes.
fname = [baseFilename '.box'];
boxFile = fullfile(pwd, fname);
vision.internal.ocr.writeBoxFile(boxFile, trainingData(j), size(I));
[~, iname] = fileparts(trainingImages(i).ImageLocation{j});
fprintf('Created box file, %s, for image %s \n', fname, iname);
end
end
end
end
end