计算机科学
弹丸
人工智能
上下文图像分类
图像(数学)
模式识别(心理学)
计算机视觉
材料科学
冶金
标识
DOI:10.1109/iceace60673.2023.10441798
摘要
Few-shot learning aims to solve the problem of model training in scenarios with scarce data. In this paper, a small sample image classification model based on contrastive learning and self-attention is proposed. This method designs a supervised contrastive loss for the pre-training process of the model, which effectively improves the model's generalization performance. In addition, this paper proposes the use of feature maps to measure the similarity between images and utilizes self-attention mechanisms to match image patches, ensuring that the model accurately compares image regions with similar relationships and makes more accurate predictions. Experimental results demonstrate that these two improvements effectively enhance the performance of small sample classification models.
科研通智能强力驱动
Strongly Powered by AbleSci AI