SAMFrun_tracker代买解读（matlab版）

万事云

于 2021-08-01 16:08:05 发布

阅读量485

点赞数

分类专栏： samf代码解读

本文链接：https://blog.csdn.net/CloudVans/article/details/119297121

版权

samf代码解读专栏收录该内容

3 篇文章 0 订阅

订阅专栏

function [precision, fps] = run_tracker(video, kernel_type, feature_type, show_visualization, show_plots)
base_path ='D:\samf data';

base_path =’ '; 引号内是样本储存的地址

if nargin < 1, video = 'choose'; end
if nargin < 2, kernel_type = 'gaussian'; end     
if nargin < 3, feature_type = 'hogcolor'; end
if nargin < 4, show_visualization = ~strcmp(video, 'all'); end  %比较strcmp中的量，相同就是1，否就是0
if nargin < 5, show_plots = ~strcmp(video, 'all'); end

nargin是输入的参数值的个数
比较strcmp中的量，相同就是1，否就是0

kernel.type = kernel_type;
features.gray = false;
features.hog = false;          
features.hogcolor = false; 

padding = 1.5;  %extra area surrounding the target 目标周围的额外区域
lambda = 1e-4;  %regularization  正则化系数λ
output_sigma_factor = 0.1;  %spatial bandwidth (proportional to target) 空间带宽（与目标成正比）

以上两段代码片是默认的设置

switch feature_type
	case 'gray'                       %灰度
		interp_factor = 0.075;  % 时间更新系数（自适应线性插值因子）

		kernel.sigma = 0.2;  %高斯核带宽
		
		kernel.poly_a = 1;  %多项式核可加项
		kernel.poly_b = 7;  %多项式核指数
	
		features.gray = true;
		cell_size = 1;
		
	case 'hog'                      %梯度直方图
		interp_factor = 0.015;
		
		kernel.sigma = 0.5;
		
		kernel.poly_a = 1;
		kernel.poly_b = 9;
		
		features.hog = true;
		features.hog_orientations = 9;
		cell_size = 4;
	case 'hogcolor'
		interp_factor = 0.01;
		
		kernel.sigma = 0.5;
		
		kernel.poly_a = 1;
		kernel.poly_b = 9;
		
		features.hogcolor = true;
		features.hog_orientations = 9;
		cell_size = 4;	
	otherwise
		error('Unknown feature.')
	end
	assert(any(strcmp(kernel_type, {'linear', 'polynomial', 'gaussian'})), 'Unknown kernel.')
    %assert判断一个表达式是否成立

switch video
case 'choose'
  video = choose_video(base_path);
        if ~isempty(video)  
                [precision, fps] = run_tracker(video, kernel_type, ...
				feature_type, show_visualization, show_plots);
			
			if nargout == 0                           %不要输出精度作为参数
				clear precision 
			end
		end
case 'all'
dirs = dir(base_path);   %获得指定文件夹下所有的子文件和文件
		videos = {dirs.name};
		videos(strcmp('.', videos) | strcmp('..', videos) | ...
		strcmp('anno', videos) | ~[dirs.isdir]) = [];
		videos(strcmpi('Jogging', videos)) = [];
		videos(end+1:end+2) = {'Jogging.1', 'Jogging.2'};
		all_precisions = zeros(numel(videos),1);
		if ~exist('matlabpool', 'file')           %exist用于查变量是否存在，存在输出1，否输出0
			                                                 % 没有并行工具箱，使用简单的“for”来迭代
			for k = 1:numel(videos)
				[all_precisions(k), all_fps(k)] = run_tracker(videos{k}, ...
					kernel_type, feature_type, show_visualization, show_plots);
			end
		else
			%并行评估所有视频的跟踪器
			if parpool('size') == 0              %parpool 是并行计算
				parpool open;
			end
			parfor k = 1:numel(videos)          %一个并行的for循环
				[all_precisions(k), all_fps(k)] = run_tracker(videos{k}, ...
					kernel_type, feature_type, show_visualization, show_plots);
			end
		end
	mean_precision = mean(all_precisions); %mean是用来计算平均值的
	fps = mean(all_fps);
	fprintf('\nAverage precision (20px):% 1.3f, Average FPS:% 4.2f\n\n', mean_precision, fps)
	if nargout > 0                                    %nargout代表输出的参数
			precision = mean_precision;
		end

isempty判断里面是否为空集，空集输出为0，非空输出为1

case 'benchmark'
		%running in benchmark mode - this is meant to interface easily
		%with the benchmark's code. 在基准测试模式下运行 - 这意味着可以轻松地与基准测试代码进行交互。
		
		%get information (image file names, initial position, etc) from
		%the benchmark's workspace variables 从基准的工作区变量中获取信息（图像文件名、初始位置等
		seq = evalin('base', 'subS'); %evalin函数用于引入工作区变量base是工作空间，subs是表达式
		target_sz = seq.init_rect(1,[4,3]);
		pos = seq.init_rect(1,[2,1]) + floor(target_sz/2); % y = floor(x) 函数将x中元素取整，值y为不大于本身的最小整数。
                                                           % 对于复数，分别对实部和虚部取整
		img_files = seq.s_frames;                          %序列帧
		video_path = []; %用来储存矩阵或者向量
		
		%call tracker function with all the relevant parameters 使用所有相关参数调用跟踪器函数
		[positions,rect_results,t]= tracker(video_path, img_files, pos, target_sz, ...
			padding, kernel, lambda, output_sigma_factor, interp_factor, ...
			cell_size, features, 0);
		
		%return results to benchmark, in a workspace variable 在工作区变量中将结果返回到基准测试
		rects =rect_results;
%         [positions(:,2) - target_sz(2)/2, positions(:,1) - target_sz(1)/2];
% 		rects(:,3) = target_sz(2);
% 		rects(:,4) = target_sz(1);
		res.type = 'rect';
		res.res = rects;
		assignin('base', 'res', res); %base是工作空间，res是变量名，res是新的值
        %assignin与evalin可以实现不同m文件主函数与子函数的工作空间变量的共存
		
		
	otherwise
		%we were given the name of a single video to process.
	    %我们得到了要处理的单个视频的名称。
		%get image file names, initial state, and ground truth for evaluation
        %获取图像文件名、初始状态和用于评估的真实情况
		[img_files, pos, target_sz, ground_truth, video_path] = load_video_info(base_path, video);
		
		
		%call tracker function with all the relevant parameters
        %使用所有相关参数调用跟踪器函数
		[positions,~, time] = tracker(video_path, img_files, pos, target_sz, ...
			padding, kernel, lambda, output_sigma_factor, interp_factor, ...
			cell_size, features, show_visualization);
		
		
		%calculate and show precision plot, as well as frames-per-second
        %计算并显示精度图，以及每秒帧数
		precisions = precision_plot(positions, ground_truth, video, show_plots);
		fps = numel(img_files) / time; %numel用于给出图像的像素数

		fprintf('%12s - Precision (20px):% 1.3f, FPS:% 4.2f\n', video, precisions(20), fps)
        %fprintf可以将数据写入指定的文本文件中

		if nargout > 0
			%return precisions at a 20 pixels threshold  以 20 像素阈值返回精度
			precision = precisions(20);
		end

	end
end