Improve Small-Sample Learning in Image Recognition Through Meta-Learning Techniques

The role of small-sample learning in improving image recognition is vital, but there is a problem that the recognition results are inaccurate. Single-sample learning cannot solve and improve the small-sample learning problem in image recognition, and the accuracy rate is low. Therefore, this paper proposes a meta-learning technique for small-sample learning analysis with expanded sample numbers. Firstly, the pixels of image recognition are used to test the small-sample learning, and the indicators are divided according to the requirements of small-sample learning to reduce the small number of samples in small-sample learning disturbing factors. Then, the influence of image recognition pixels on small-sample learning is formed to form a small-sample learning scheme, and the accuracy of small-sample learning is comprehensively analyzed. MATLAB simulation shows that under specific evaluation criteria, the amount of information extracted and the accuracy of sample extraction by meta-learning technology for small-sample learning is better than single-sample learning.