matlab-使用matlab实现基于PointNet++网络的点云数据分类仿真分析

classdef PtCloudClassificationDatastore < matlab.io.Datastore % PtCloudClassificationDatastore Datastore with point clouds % and associated categorical labels. % % ds = PtCloudClassificationDatastore(foldername) constructs a datastore % that represents point clouds and associated categories for your dataset. % The input, foldername, is a string or char array % which represents the name of the folder that stores trining or test point cloud data. properties MiniBatchSize = 1; end properties(Dependent) Files end properties(Access = private) FileDatastore end methods function this = PtCloudClassificationDatastore(datapath) % Please put the training and test data folder in your current % directly. if nargin==0 return; end this.FileDatastore = fileDatastore(datapath,'ReadFcn',@extractTrainingData, ... 'IncludeSubfolders',true,'FileExtensions',{'.ply','.pcd'}); end function tf = hasdata(this) tf = hasdata(this.FileDatastore); end function [data,info] = read(this) if ~hasdata(this) error('Reached end of data. Reset datastore.'); end % Preallocate output. batchSize = this.MiniBatchSize; data = cell(batchSize,2); info = struct(... 'Filename',cell(batchSize,1),... 'FileSize',cell(batchSize,1)); % Read mini-batch size worth of data. The size of data can be % less than the specified batch size. idx = 0; while hasdata(this.FileDatastore) idx = idx + 1; [data(idx,:),info(idx)] = read(this.FileDatastore); if idx == batchSize break; end end data = data(1:idx,:); info = info(1:idx); end function reset(this) reset(this.FileDatastore); end function files = get.Files(this) files = this.FileDatastore.Files; end function set.Files(this,files) this.FileDatastore.Files = files; end end methods(Access=protected) function newds = copyElement(this) newds = PtCloudClassificationDatastore(); newds.FileDatastore = copy(this.FileDatastore); newds.MiniBatchSize = this.MiniBatchSize; % Shuffle files and corresponding labels numObservations = numel(newds.Files); idx = randperm(numObservations); newds.Files = newds.Files(idx); end end end function dataOut = extractTrainingData(fname) pointData = readbin(fname); % The label information was obtained using the folder name idx=strfind(fname,'\'); name=fname(idx(end-1)+1:idx(end)-1); dataOut = {pointCloud(pointData),categorical(cellstr(name))}; end function pointData = readbin(fname) % readbin Read point in the trainig or test files. pointData = pcread(fname); pointData =pointData.Location; end