How does Google's image search work?

Answer 1

Google is constantly running a computer called a 'crawler' or 'bot' that constantly roams the Internet and it's 3,000,000,000 webpages (growing by the min)... When it comes across a new image, it stores its location on the Internet and links the title of the image to specific keywords. When search terms matching the keywords are input by a user on the Google Images web site, Google calls up all of the images that have been stored by the 'crawler'.
The 'crawler' is a MASSIVE supercomputer doing MILLIONS of calculations a second, storing, searching, storing, searching...

Answer 2

By hand,
Google researchers visited a lot of sites, looked at their pictures and put a note (or tag) in their database on the picture linked to the website. The tag is one that can be searched for in database style.

Computers push numbers they don't really understand anything. That requires intelligence. We can build forms of artificial intelligence, but they are fairly primitive. Object recognition is very difficult for a computer to do (and robots are just computers with arms). So far we can get a computer to identify something simple like a cup, but more complicated items like who is the celebrity in a picture of 3 people are just beyond them.

So all of the image classification has to be done by individuals who will then make very short notes that can be put into search-able databases.

This summer the magazine Popular Science wrote an article naming the top 10 most inventive new scientists in their field. One of these people was a man who invented a game where the players are shown various pictures and they write a few notes about them or give them image tags which can be put into a search-able database and linked to the sites where these images came from. As of the publication of that article Google has not shown any interest in developing or using his idea.

So for now all image indexing has to be done by humans. They go to a website and examine some of the pictures there, then write some tag lines that can be put into a search-able database linked to the website where the picture came from. All of these tags are put into Goggle’s Image Database for you to search through.

Web Crawlers and spiders search sites for TEXT and automatically index it. If a picture is labeled then that TEXT can be found and indexed. But, if the picture has the file name of housepic01 then the spider has no idea what that means or what the picture is, so it can't index them. Fan websites that offer pictures of a celebrity are easy to index, only the website's homepage needs to be noted with the celebrities name as the tag. Websites about a subject usually contain an image of that subject, so it is easy to make tags for these sites. If you search for pictures of houses or rocks then someone had to go to those websites to make sure they had those pictures. Until someone actually indexes the websites it is impossible for Google to know what the pictures on a website actually are.

Answer 3

I don't know about Google Images, but I know if you search for images on Yahoo without content filters, then just about every search term you can think of will come back with some kind of porn.

Answer 4

just click on image and type what u want to see