Caltech Pedestrian Detection Benchmark

传送门：http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/

环境：ubuntu，matlab

 下载toolbox及评估代码

#!/bin/bash

# download toolbox
git clone https://github.com/pdollar/toolbox pdollar_toolbox

# download evaluation/labeling code
wget http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/code/code3.2.1.zip
mkdir code
unzip code3.2.1.zip -d code
rm code3.2.1.zip

#download annotations
cd code
mkdir data-USA
cd data-USA
wget http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/datasets/USA/annotations.zip
unzip annotations.zip
rm annotations.zip

#download results
mkdir res
cd  res
list=("VJ" "HOG" "FPDW" "RPN+BF" "MS-CNN" "SDS-RCNN" "TLL-TFA" "F-DNN2+SS" "F-DNN+SS" "GDFL" "PCN" "F-DNN" "AdaptFasterRCNN" "SA-FastRCNN" "UDN+" "FasterRCNN+ATT")
for method in "${list[@]}"
do
    http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/datasets/USA/res/${method}.zip
    unzip ${method}.zip
    rm ${method}.zip
done

评估自己的实验结果

一、通用步骤

1、依照官网要求的格式处理自己的数据，放到code/data-USA/res文件夹下。

官网要求的格式：res/目录下新建一个文件夹命名为自己方法的名字（如：my_method），res/my_method/set06/V000/I00029.txt，每个txt文件内存放对应图片的检测结果，每行的格式为[left, top, width, height]

2、用matlab评估

# 第一次使用
p = genpath('../pdollar_toolbox');
addpath(p);


#画MR-FPPI图
将算法结果放入./data-USA/res/XXX/
修改dbEval.m文件，exps只保留Reasonble，变量algs只保留你需要的方法、然后加上自己的方法，变量dataNames只保留UsaTest

最终生成的结果会存放到code/results文件夹下

二、YOLOv3检测行人的结果

exps = Reasonble时，MR=11%，在所有的方法里面排在第13名。

这个结果是博主自己用yolov3只检测行人，然后在评估代码上跑出来的。

如果需要的可以在这里下载：https://download.csdn.net/download/qq_33614902/10659040

将下载后的文件解压放到data-USA/res/文件夹下，然后如下修改dbEval.m文件，最后运行dbEval.m

function dbEval
% Evaluate and plot all pedestrian detection results.
%
% Set parameters by altering this function directly.
%
% USAGE
%  dbEval
%
% INPUTS
%
% OUTPUTS
%
% EXAMPLE
%  dbEval
%
% See also bbGt, dbInfo
%
% Caltech Pedestrian Dataset     Version 3.2.1
% Copyright 2014 Piotr Dollar.  [pdollar-at-gmail.com]
% Licensed under the Simplified BSD License [see external/bsd.txt]

% List of experiment settings: { name, hr, vr, ar, overlap, filter }
%  name     - experiment name
%  hr       - height range to test
%  vr       - visibility range to test
%  ar       - aspect ratio range to test
%  overlap  - overlap threshold for evaluation
%  filter   - expanded filtering (see 3.3 in PAMI11)
exps = {
  'Reasonable',     [50 inf],  [.65 inf], 0,   .5,  1.25
   'All',            [20 inf],  [.2 inf],  0,   .5,  1.25
%   'Scale=large',    [100 inf], [inf inf], 0,   .5,  1.25
   'Scale=near',     [80 inf],  [inf inf], 0,   .5,  1.25
   'Scale=medium',   [30 80],   [inf inf], 0,   .5,  1.25
   'Scale=far',      [20 30],   [inf inf], 0,   .5,  1.25
   'Occ=none',       [50 inf],  [inf inf], 0,   .5,  1.25
   'Occ=partial',    [50 inf],  [.65 1],   0,   .5,  1.25
   'Occ=heavy',      [50 inf],  [.2 .65],  0,   .5,  1.25
%   'Ar=all',         [50 inf],  [inf inf], 0,   .5,  1.25
   'Ar=typical',     [50 inf],  [inf inf],  .1, .5,  1.25
   'Ar=atypical',    [50 inf],  [inf inf], -.1, .5,  1.25
%   'Overlap=25',     [50 inf],  [.65 inf], 0,   .25, 1.25
%   'Overlap=50',     [50 inf],  [.65 inf], 0,   .50, 1.25
%   'Overlap=75',     [50 inf],  [.65 inf], 0,   .75, 1.25
%   'Expand=100',     [50 inf],  [.65 inf], 0,   .5,  1.00
%   'Expand=125',     [50 inf],  [.65 inf], 0,   .5,  1.25
%   'Expand=150',     [50 inf],  [.65 inf], 0,   .5,  1.50 
};
exps=cell2struct(exps',{'name','hr','vr','ar','overlap','filter'});

% List of algorithms: { name, resize, color, style }
%  name     - algorithm name (defines data location)
%  resize   - if true rescale height of each box by 100/128
%  color    - algorithm plot color
%  style    - algorithm plot linestyle
n=1000; clrs=zeros(n,3);
for i=1:n, clrs(i,:)=max(.3,mod([78 121 42]*(i+1),255)/255); end
algs = {
  'VJ',               0, clrs(1,:),   '-'
  'HOG',              1, clrs(2,:),   '--'
%   'FtrMine',          1, clrs(3,:),   '-'
%   'Shapelet',         0, clrs(4,:),   '--'
%   'PoseInv',          1, clrs(5,:),   '-'
%   'MultiFtr',         0, clrs(6,:),   '--'
%   'MultiFtr+CSS',     0, clrs(7,:),   '-'
%   'MultiFtr+Motion',  0, clrs(8,:),   '--'
%   'HikSvm',           1, clrs(9,:),   '-'
%   'Pls',              0, clrs(10,:),  '--'
%   'HogLbp',           0, clrs(11,:),  '-'
%   'LatSvm-V1',        0, clrs(12,:),  '--'
%   'LatSvm-V2',        0, clrs(13,:),  '-'
%   'ChnFtrs',          0, clrs(14,:),