The Multi-Angle Recognition Engine (MARE) engine equipped in the application is able to reconstruct the 3D model of insects through a single photo, breaking the limitation that traditional recognition requires a standard side view. The core of the technology is the topological relationship analysis of 72 key feature points, and even if only a partial part of the insect's wing (≥30% area) is photographed, a recognition accuracy of 89% can still be achieved.
The innovative "Feature Completion Algorithm" can infer the back pattern according to the direction of the markings, and the success rate of recognizing mutilated specimens is 43% higher than that of Google Lens.The system will also guide the shooting intelligently: when it detects a poor angle, the AR interface will mark the suggested shooting direction.
The technology, for which seven patents have been applied, has been used in forestry pest monitoring, increasing the efficiency of field workers by 2.7 times. This year, it will be upgraded to a multi-spectral recognition version, with the added function of reading age through wing vein patterns.
This answer comes from the articleBugPic: a tool for quickly identifying insects by taking photosThe
English 
简体中文 
日本語 
Deutsch 
Português do Brasil