ALIPRHELP

The ALIPR (pronounced a-lip-er) is a machine learning based image tagging and searching system. The system learns from human annotations of images, builds compact mathematical knowledge bases about the concepts, and suggests tags to unseen images. Unlike many other image search engines which rely on file names or surrounding text to search for images, the ALIPR tagging function makes use of no information other than the actual pixel information within the pictures.

Image Uploading and Tagging

On the front page of the Website, you can upload an image file of yours from your computer. Click on the "Browse" button to find the image file on your computer. Then click on "alipr it!" to let ALIPR take a look at your pictures. After about one second, ALIPR gets the picture from your computer, processes it mathematically, compares it with the mathematical knowledge base of all the hundreds of concepts in the "brain" of ALIPR, and selects 15 most likely tags to describe the content of your picture. We would like you to help check the correctness of these tags by checking the good predicted tags. If you can think of any missing tags, add them in the provided field. See below for an example. Finally, click on "teach alipr", your pictures in then entered in a keyword-searchable collection and will be used to teach ALIPR as she grows up. Our goal is to build a highly useful ALIPR image tagging assistant to benefit the human beings for generations to come.



If you find some interesting pictures on the Web, or if your pictures are on a Website already, you can use the URL uploading function on the front page of ALIPR.com. For certain browsers, including firefox, you can simply drag and drop from another browser showing the picture. Then follow the same steps as above to check off the correct tags and add missing tags. This way, the picture will be available for keyword-based searching and be used for teaching ALIPR.


Keyword Search



You can search images uploaded by users using keywords. The keywords used in the search include those ALIPR-predicted tags check by users and those keyword tags added by the users. For example, you can search with one keyword like "motorcycle" or a combination of keywords like "mountain lake". You can also click on suggested keyword.



If you just want to have a look at images randomly, you can click on "Random" or "Recent" links on the pages to get a random collection, or a recent collection of images uploaded.

Related and Visual Similarity Search

You can search images uploaded by users using an image as a query. The visual similarity search is powered by the SIMPLIcity search engine we developed in the past.


Related Image Search


Visual Similarity Search

Voting

You can vote for images you like. When you get into the Keyword Search page, without a specific search keyword, the page shows a random set of images from the database. If you like any one of them, simply click the circle below the image to vote. If you don't like any of them, simply click "next" to get a new set of random images. At any point you would like to start keyword search, just type the keyword in the search field, or click on any of the suggested keyword.



© alipr.com