Skip to content

ysix7/Dataset

Repository files navigation

常用数据集1

常用数据集2

BSDS500/300

这两个是berkely (伯克利大学)computer vision group提供的数据集,用来做segmentation**(图像分割)或contour detection(轮廓检测)的,还有人拿这个做super resolution(超分重建**)的

数据库包含200张训练图,200张侧视图,100张检验图。ground truth是人工标识的,已数据库图片id为单位,保存成mat格式文件,一个文件包含多个标记者的标记信息。有轮廓和分割信息,用matlab读取很方便,直接load就行。

下载地址

数据库下载链接

读取脚本

生成轮廓可视化图:

% make_gt_bondary_image.m
% bsdsRoot 需包含train、test 两个文件夹,生成的结果也放在这两个文件夹的bon子文件夹里
state = 'test';%修改为test或train,分别处理两个文件夹 
file_list = dir(fullfile(bsdsRoot,state,'*.mat'));%获取该文件夹中所有jpg格式的图像
for i=1:length(file_list)
    mat = load(file_list(i).name);
    [~,image_name,~] = fileparts(file_list(i).name);
    gt = mat.groundTruth;
    for gtid=1:length(gt)
        bmap = gt{gtid}.Boundaries;
        if gtid==1
            image = bmap;
        else
            image(bmap==true)=true;
        end

    end
    %黑底白边
    imwrite(double(image),fullfile(bsdsRoot,'bon',state,[image_name '.jpg']));
    %白底黑边
    %imwrite(1-double(image),fullfile(bsdsRoot,state,'bon',[image_name '.jpg']));

end

生成分块可视化:

% make_gt_seg_image.m
% bsdsRoot 需包含train、test 两个文件夹,生成的结果也放在这两个文件夹的seg子文件夹里
state = 'train';
file_list = dir(fullfile(bsdsRoot,state,'*.mat'));%获取该文件夹中所有jpg格式的图像
for i=1:length(file_list)
    mat = load(file_list(i).name);
    [~,image_name,~] = fileparts(file_list(i).name);
    gt = mat.groundTruth;
    for gtid=1:length(gt)
        seg = double(gt{gtid}.Segmentation);
        seg = seg/max(max(seg));
        if gtid == 1
            image = seg;
        else
            image = image+seg;
        end
    end
    image = image/length(gt);
    imwrite(double(image),fullfile(bsdsRoot,state,'seg',[image_name '.jpg']));

end

BSD68

Color BSD68 dataset for image denoising benchmarks It is part of The Berkeley Segmentation Dataset and Benchmark. This benchmark dataset is widely used for measuring image denoising algorithms performance.

Set12

#CBSD68

下载地址

About

datasets for experiments

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages