Handling images has always posed serious challenges to search engines. Search engines are not as smart as human brain and they cannot recognize images at least so far. They have to go by the tags that are associated to the image. Computers need text tags to recognize images. However, one of the papers presented by Google this Monday in Miami can disagree with this view. Google is conducting a research in computer vision. As per the latest breakthrough made by Google in image recognition capabilities, the new technology will empower computers in recognizing over 50000 landmarks from the images. However, the question is how accurate are these image recognitions? Google responds saying the image recognition technology is currently producing 80% accuracy. Though it might sound that we have progressed a great deal in this area, 20% of grey area for image recognition is still a large gap to be filled. Its usage may not be extended to face recognition as yet.
The pattern recognition or image recognition project still under the research and they are yet to cover new grounds in this territory. While the entire world is still struggling to handle the images, Google’s initiative has gone a considerable distance in handling the issue at hand, which certainly has to be recognized and appreciated.
Google will use in the initial stage around 40 million photos that are tagged in GPS from its image services Picasa and Panoramio. Google will also use web pages on online tour guides. A complete list of all popular landmarks will be compiled from this collection of images. Amongst this vast collection of images, a candidate image will be used for each landmark and the images will be pruned using latest technologies and image matching techniques.
The next crucial phase indexing the massive volume of images so that the images can be recognized by the search engines fast. Will the new images too be recognized is still a question. There is lot of room for confusion because same architecture can be shared by multiple landmarks that are in no way linked. This makes closing the grey area all the more difficult. While we are trying various ways to recognize images, we suddenly realize the brilliance of human brain, which can match images and differentiate millions of objects. To surpass the cleverness of the human brain we need another million years and that too is doubtful that we will ever succeed. What we are trying to do in the case of image recognition, just a glimpse of what our human brains can do with ease.
Google hopes to enhance its Computer Vision project results by increasing the accuracy rate of image recognition.