📄 faces.html~
字号:
<h1>Face detection</h1>
These are some quick experiemnts on face detection, the code was written from scratch in about 2 hours, using this group photo to test:
<p>
<img src="group2.jpg">
<p>
One of the guys on the back row (4th from the right) is cut out as an exemplar template. A crude double-or loop is used to search for the template at all horizontal and vertical positions (no scaling or rotation). It takes about 5 minutes to run in matlab. Template checking is done by normalising the histogram means and variences for each candidate area, and for the template. A threshold output is used, and the active areas are reduced to points with erosion. The output looks like this:
<p>
<img src="out.bmp">
<p>
Note it doesnt do too well on bald heads, beards, long hair, people standing sideways-on, and people at the front, whose scales are different.
<p>
The only fiddle parameter used was the activation threshold.
<p>
The matlab code is available, contact me if you're interested.
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -