Scientists at Brigham Young University (BYU) have developed an algorithm that can accurately identify objects in images or videos and can learn to recognize new objects on its own.
Although other object recognition systems exist, the Evolution-Constructed Features algorithm is notable in that it decides for itself what features of an object are significant for identifying the object and is able to learn new objects without human intervention. The researchers say that unlike other methods, it does not require retuning or reworking for different tasks.
"In most cases, people are in charge of deciding what features to focus on and they then write the algorithm based off that,” says Dr. Dah-Jye Lee, professor of electrical and computer engineering at BYU and author of the paper. “With our algorithm, we give it a set of images and let the computer decide which features are important."
According to Dr. Lee, most other algorithms require a lot of fine-tuning of parameters and methods to achieve their best accuracy, whereas the Evolution-Constructed Features algorithm does not. Despite this, the researchers ay the algorithm has performed as well or better in object recognition tests than other leading object recognition algorithms.
For example, the Evolution-Constructed Features algorithm achieved 100 percent accuracy on motorbike, face, airplane and car image datasets from Caltech.
Caltech's database is used to benchmark the algorithm against other similar research, with other "published well-performing object recognition systems" scoring 95-98 percent accuracy in the same tests.
Lee and his team suggest that the algorithm could be used applications such as detecting invasive fish species or identifying flaws in produce such as apples. To this end, it was also shown to have 99.4 percent accuracy on fish species image datasets from BYU’s own biology department.
"Within some predefined criteria and situation, object recognition will continue to show its progress," predicts Lee. "Who knows? Maybe one day, when the computation power of computing platforms increases to be close to human brain, we could see some real breakthrough."
Source: Brigham Young University
