YIBADA

Wolfram’s New Image Identification Software Recognizes Photos, Makes Human Errors

| May 15, 2015 01:36 AM EDT

Wolfram Language logo

Wolfram Language re-launched its Image Identify Image Identification Project on May 14, Thursday. The image identifier software recognizes uploaded photos, and makes human-like errors.

Wolfram Research CEO Stephen Wolfram wrote in a blog entry that the image identifier robot is not perfect, but is usually correct. He also believes it performs "remarkably well," according to PC Mag.

Wolfram is amazed by the wrong answers of the image identification system because the errors are quite human-like.

The Image Identify tool can analyze any digital image, such as from a digital camera or an Internet website.  After the image identifier app guesses which object is in the picture, it then provides more details about it.

Users can then request for a definition for the ImageIdentify quiz result, or ask a Wikipedia page to create a word cloud. They can also use the image identification technology to build apps or design websites.

Wolfram explained that the Wolfram image identifier is a useful example of artificial intelligence. However, more important is the use of the AI technology for "knowledge-based programming."

An early bug in the image identification software was that it did not search for human faces. For example, a picture of Indiana Jones in his fedora produced this result: hat.

Wolfram described the process of ImageIdentify 's software debugging as an "interesting process." However, most of them have been fixed.

Wolfram's image recognition tech follows similar projects by other tech giants. They include those from Google, Amazon, and Microsoft, according to Engadget.

The Wolfram Image Identify online app was first released in March.

Related News

Most Popular

EDITOR'S PICK