Few-shot Object Recognition based on Three-Way Decision and Active Learning
Li, B., Luo, S., Wang, J., Tian, L. and Chen, D. (2022). Few-shot Object Recognition based on Three-Way Decision and Active Learning. Visual Computer . 37.
|Authors||Li, B., Luo, S., Wang, J., Tian, L. and Chen, D.|
The problem of object recognition with few-shot or a few labeled samples is a very challenging research. Focusing on the problem of few shot object recognition, it is introduced a method of object recognition method based on active learning. To solve the recognition problem of unknown classes under the condition of few shot, based on the object recognition with active learning, we propose a sample selection method based on three-way decision. Firstly, we divide the unlabeled samples into three disjoint domains according to the recognition probability and the three-way decision threshold. Then, we use different query strategies for the three domains, as far as possible to select new samples of unknown classes from unlabeled samples, enrich the recognition types of the classification model, and optimize the performance of the classification model. Then, this paper introduces an algorithm for solving three-way decision threshold. Finally, simulation experiments verify the feasibility and effectiveness of the algorithm proposed in this chapter in object recognition.
|Keywords||few-shot object recognition; three-way decision; active learning; few-shot sample selection; object recognition|
|Web address (URL)||https://www.springer.com/journal/371|
|Publication process dates|
|Accepted||20 Jan 2022|
|Deposited||31 Jan 2022|
|Accepted author manuscript|
File Access Level
1views this month
0downloads this month