An Interactive Annotation Method Based on Incremental Learning

• Published in: 2023 International Conference on Image Processing, Computer Vision and Machine Learning (ICICML) pp. 1181 - 1184
• Main Authors: Duan, Xiusheng, Sun, Guohua, Cao, Jingya, Zhang, Yuyang, Yang, Jingchao, Han, Duyu
• Format: Conference Proceeding
• Language: English
• Published: IEEE 03.11.2023
• Subjects: Computational modeling; data annotation; incremental learning; interaction; Learning systems; Machine learning; Military equipment; Optimization methods; sample augmentation; Target recognition; Training
• DOI: 10.1109/ICICML60161.2023.10424823

In recent years, military science and technology has been developed rapidly, and new equipments has been equipped in the army. Augmented Reality (AR) technology provides the possibility to solve the problem of equipments operation training. But in the training process, One of the essential technologies is how to location and identify the operation keys. Obviously, the sample set of equipments operation keys is the basis of key recognition, and a certain scale of sample set is an indispensable element to ensure the performance of the recognition model. On the basis of manually establishing a small sample set, this paper puts forward a mechanism based on interactive automatic sample labeling and sample set expanding, the target recognition model was updated by incremental learning method at the same time.