matlab 将groundTruth转为coco 格式

网上下载了模板，根据实际结构体进行了修改。

版本2021

Fonte = 'Rebar'; 

len = size(imageLabelingSession.ImageFilenames,1);%size(labelingSession.ImageSet.ImageStruct,2);


for i=1: 1%len
    
    path = imageLabelingSession.ImageFilenames{i};%labelingSession.ImageSet.ImageStruct(i).imageFilename; % resgata o endereço completo da foto
    VOC.annotation.path.Text = path;
    % Separa o texto
    expression = '\\';
    splitStr = regexp(path,expression,'split');    
    % Nome da foto
    foto = string(splitStr(size(splitStr,2)));
    VOC.annotation.filename.Text = char(foto); 
    % Localiza o diretório
    folder = string(splitStr(size(splitStr,2)-1));
    VOC.annotation.folder.Text = char(folder);  % Nome do diretório da foto
    VOC.annotation.source.database.Text = Fonte; % Define a Fonte dos dados
    % Características da Imagem
    [height, width, depth] = size(imread(path));
    VOC.annotation.size.width.Text = num2str(width);
    VOC.annotation.size.height.Text = num2str(height);
    VOC.annotation.size.depth.Text = num2str(depth);
    % Segmentação
    VOC.annotation.segmented.Text = '0'
    
    
    number_bboxes = size(labelingSession.ImageSet.ImageStruct(i).objectBoundingBoxes,1);
    
    for j=1:number_bboxes    
              
        % Objetos - Boundboxes
            %Nome da Tag
        categoryID = labelingSession.ImageSet.ImageStruct(i).catID(j);
        VOC.annotation.object{1, j}.name.Text = labelingSession.CategorySet.CategoryStruct(categoryID).categoryName;
        VOC.annotation.object{1, j}.pose.Text = 'Unspecified'; 
        VOC.annotation.object{1, j}.truncated.Text = '0';
        VOC.annotation.object{1, j}.difficult.Text = '0';
        % Posição da BoundBox  [x,y,width, height]
        xmin = labelingSession.ImageSet.ImageStruct(i).objectBoundingBoxes(j); 
        ymin = labelingSession.ImageSet.ImageStruct(i).objectBoundingBoxes(j+ number_bboxes);
        xmax = xmin + labelingSession.ImageSet.ImageStruct(i).objectBoundingBoxes(j+(2*number_bboxes));
        ymax = ymin + labelingSession.ImageSet.ImageStruct(i).objectBoundingBoxes(j+(3*number_bboxes));
        VOC.annotation.object{1, j}.bndbox.xmin.Text = xmin;
        VOC.annotation.object{1, j}.bndbox.ymin.Text = ymin;
        VOC.annotation.object{1, j}.bndbox.xmax.Text = xmax;
        VOC.annotation.object{1, j}.bndbox.ymax.Text = ymax;
                
    end
    
    expression = '\.';
    foto1 = regexp(foto,expression,'split');
    nome = sprintf('%s.xml',foto1(1));
    struct2xml(VOC,nome);
    clear VOC;
    
    
end

结构体

function varargout = struct2xml( s, varargin )
%Convert a MATLAB structure into a xml file 
% [ ] = struct2xml( s, file )
% xml = struct2xml( s )
%
% A structure containing:
% s.XMLname.Attributes.attrib1 = "Some value";
% s.XMLname.Element.Text = "Some text";
% s.XMLname.DifferentElement{1}.Attributes.attrib2 = "2";
% s.XMLname.DifferentElement{1}.Text = "Some more text";
% s.XMLname.DifferentElement{2}.Attributes.attrib3 = "2";
% s.XMLname.DifferentElement{2}.Attributes.attrib4 = "1";
% s.XMLname.DifferentElement{2}.Text = "Even more text";
%
% Will produce:
% <XMLname attrib1="Some value">
%   <Element>Some text</Element>
%   <DifferentElement attrib2="2">Some more text</Element>
%   <DifferentElement attrib3="2" attrib4="1">Even more text</DifferentElement>
% </XMLname>
%
% Please note that the following strings are substituted
% '_dash_' by '-', '_colon_' by ':' and '_dot_' by '.'
%
% Written by W. Falkena, ASTI, TUDelft, 27-08-2010
% On-screen output functionality added by P. Orth, 01-12-2010
% Multiple space to single space conversion adapted for speed by T. Lohuis, 11-04-2011
% Val2str subfunction bugfix by H. Gsenger, 19-9-2011
    
    if (nargin ~= 2)
        if(nargout ~= 1 || nargin ~= 1)
            error(['Supported function calls:' sprintf('\n')...
                   '[ ] = struct2xml( s, file )' sprintf('\n')...
                   'xml = struct2xml( s )']);
        end
    end

    if(nargin == 2)
        file = varargin{1};

        if (isempty(file))
            error('Filename can not be empty');
        end

        if (isempty(strfind(file,'.xml')))
            file = [file '.xml'];
        end
    end
    
    if (~isstruct(s))
        error([inputname(1) ' is not a structure']);
    end
    
    if (length(fieldnames(s)) > 1)
        error(['Error processing the structure:' sprintf('\n') 'There should be a single field in the main structure.']);
    end
    xmlname = fieldnames(s);
    xmlname = xmlname{1};
    
    %substitute special characters
    xmlname_sc = xmlname;
    xmlname_sc = strrep(xmlname_sc,'_dash_','-');
    xmlname_sc = strrep(xmlname_sc,'_colon_',':');
    xmlname_sc = strrep(xmlname_sc,'_dot_','.');

    %create xml structure
    docNode = com.mathworks.xml.XMLUtils.createDocument(xmlname_sc);

    %process the rootnode
    docRootNode = docNode.getDocumentElement;

    %append childs
    parseStruct(s.(xmlname),docNode,docRootNode,[inputname(1) '.' xmlname '.']);

    if(nargout == 0)
        %save xml file
        xmlwrite(file,docNode);
    else
        varargout{1} = xmlwrite(docNode);
    end  
end

% ----- Subfunction parseStruct -----
function [] = parseStruct(s,docNode,curNode,pName)
    
    fnames = fieldnames(s);
    for i = 1:length(fnames)
        curfield = fnames{i};
        
        %substitute special characters
        curfield_sc = curfield;
        curfield_sc = strrep(curfield_sc,'_dash_','-');
        curfield_sc = strrep(curfield_sc,'_colon_',':');
        curfield_sc = strrep(curfield_sc,'_dot_','.');
        
        if (strcmp(curfield,'Attributes'))
            %Attribute data
            if (isstruct(s.(curfield)))
                attr_names = fieldnames(s.Attributes);
                for a = 1:length(attr_names)
                    cur_attr = attr_names{a};
                    
                    %substitute special characters
                    cur_attr_sc = cur_attr;
                    cur_attr_sc = strrep(cur_attr_sc,'_dash_','-');
                    cur_attr_sc = strrep(cur_attr_sc,'_colon_',':');
                    cur_attr_sc = strrep(cur_attr_sc,'_dot_','.');
                    
                    [cur_str,succes] = val2str(s.Attributes.(cur_attr));
                    if (succes)
                        curNode.setAttribute(cur_attr_sc,cur_str);
                    else
                        disp(['Warning. The text in ' pName curfield '.' cur_attr ' could not be processed.']);
                    end
                end
            else
                disp(['Warning. The attributes in ' pName curfield ' could not be processed.']);
                disp(['The correct syntax is: ' pName curfield '.attribute_name = ''Some text''.']);
            end
        elseif (strcmp(curfield,'Text'))
            %Text data
            [txt,succes] = val2str(s.Text);
            if (succes)
                curNode.appendChild(docNode.createTextNode(txt));
            else
                disp(['Warning. The text in ' pName curfield ' could not be processed.']);
            end
        else
            %Sub-element
            if (isstruct(s.(curfield)))
                %single element
                curElement = docNode.createElement(curfield_sc);
                curNode.appendChild(curElement);
                parseStruct(s.(curfield),docNode,curElement,[pName curfield '.'])
            elseif (iscell(s.(curfield)))
                %multiple elements
                for c = 1:length(s.(curfield))
                    curElement = docNode.createElement(curfield_sc);
                    curNode.appendChild(curElement);
                    if (isstruct(s.(curfield){c}))
                        parseStruct(s.(curfield){c},docNode,curElement,[pName curfield '{' num2str(c) '}.'])
                    else
                        disp(['Warning. The cell ' pName curfield '{' num2str(c) '} could not be processed, since it contains no structure.']);
                    end
                end
            else
                %eventhough the fieldname is not text, the field could
                %contain text. Create a new element and use this text
                curElement = docNode.createElement(curfield_sc);
                curNode.appendChild(curElement);
                [txt,succes] = val2str(s.(curfield));
                if (succes)
                    curElement.appendChild(docNode.createTextNode(txt));
                else
                    disp(['Warning. The text in ' pName curfield ' could not be processed.']);
                end
            end
        end
    end
end

%----- Subfunction val2str -----
function [str,succes] = val2str(val)
    
    succes = true;
    str = [];
    
    if (isempty(val))
        return; %bugfix from H. Gsenger
    elseif (ischar(val))
        %do nothing
    elseif (isnumeric(val))
        val = num2str(val);
    else
        succes = false;
    end
    
    if (ischar(val))
        %add line breaks to all lines except the last (for multiline strings)
        lines = size(val,1);
        val = [val char(sprintf('\n')*[ones(lines-1,1);0])];
        
        %transpose is required since indexing (i.e., val(nonspace) or val(:)) produces a 1-D vector. 
        %This should be row based (line based) and not column based.
        valt = val';
        
        remove_multiple_white_spaces = true;
        if (remove_multiple_white_spaces)
            %remove multiple white spaces using isspace, suggestion of T. Lohuis
            whitespace = isspace(val);
            nonspace = (whitespace + [zeros(lines,1) whitespace(:,1:end-1)])~=2;
            nonspace(:,end) = [ones(lines-1,1);0]; %make sure line breaks stay intact
            str = valt(nonspace');
        else
            str = valt(:);
        end
    end
end

效果