Training a high-quality image classification model often requires a large number of labeled datasets. However, in real-world business scenarios or production environments, the cost of obtaining labeled samples can be prohibitively high. Therefore, finding ways to obtain valuable labeled data at a lower cost is essential to improve the effectiveness of the algorithm. In this paper, we propose an effective and novel image selection strategy for active learning called Competence-based Active Learning. Our approach selects informative source images to strengthen the model’s generalization ability while minimizing the cost of manual labeling. Our experiments on MNIST, FashionMNIST, CIFAR10, and CIFAR100 demonstrate that our selection approach outperforms random selection and conventional selection methods. Moreover, Competence-based Active Learning not only enhances the generalization ability of the model but also reduces the training time.