Dah - Jye Lee said, in theory, it is a process similar to the teacher to teach children to learn. In this study, Li Dajie using a load of various image sample set (he picked out the face samples, airplanes, cars and motorcycles, etc) of the computer, and the new algorithm mentioned above (known as "ECO the features"), and then let your computer you find object recognition important elements.
For example, the computer may find unique Angle between aircraft fuselage and wings, and then to identify the difference between planes and cars. Li Dajie's team found that the new algorithm in the recognition of four groups of data sets, each set of accuracy reached 100%.
In addition, although more difficult to identify a variety of different objects, but the ECO features algorithm can still keep a higher accuracy. Li Dajie algorithm on four species of fish, which can identify accuracy can reach 99.4%. Similar object recognition systems, in contrast, in the judgment when difference of the four species of fish the best accuracy can reach 98%. Therefore, in the same type of object recognition, ECO the features to win again.
Dah - Jye Lee team think, ECO the features in unmanned or human intensive situation could have many USES, such as tracking a habitat of invasive species, lock line some products for defects. In addition, this algorithm has huge potential applications in the Internet of things, in such an environment, can let the system monitoring in and out of the structure of the individuals, tracking user food left in the refrigerator, and even help drive vehicles in the smart city and so on.