基于多目标乌鸦搜索算法优化DenseNet图像分类算法研究
Optimization of DenseNet Image Classification Algorithm Based on Multi-objective Crow Search Algorithm
胡容俊 1王正红1
作者信息
- 1. 北华大学计算机科学技术学院,吉林吉林 132000
- 折叠
摘要
图像分类是计算机视觉领域中的关键任务,其目标是将输入的图像分配到预定义的类别中,核心思想是通过学习从图像的局部特征中提取高级抽象表示,使模型能够有效识别并区分不同类别的图像.图像分类已应用于诸多领域,包括医学影像识别、自动驾驶、安全监控等.但图像分类也存在一些问题,如小样本问题、类别不平衡及对抗攻击等.近年来,随着深度学习的迅速发展,卷积神经网络(CNN)在图像分类任务中取得了显著的效果.设计了一种启发式算法,引入多目标乌鸦搜索算法,解决多目标优化问题,通过实验,与其他先进算法进行比较,验证了优化后的DenseNet在图像分类任务上性能有所提升,可优化卷积神经网络模型在图像分类中的应用.
Abstract
Image classification is a crucial task in computer vision,aiming to assign input images to predefined categories.The core idea of this task lies in learning high-level abstract representations extracted from local features of images,enabling the model to effectively identify and differentiate images across various categories.Image classification holds significant importance in numerous application domains,including medical image recognition,autonomous driving,and security surveillance.However,the field of image classification faces challenges such as the small sample problem,class imbalance,and adversarial attacks,etc.In recent years,with the rapid development of deep learning,Convolutional Neural Networks(CNNs)have achieved remarkable success in image classification tasks.Considering these factors,designing a heuristic algorithm to optimize the application of CNNs in image classification holds substantial value.
关键词
多目标乌鸦搜索算法/密集卷积网络/特征提取/深度学习/图像分类Key words
Multi-objective crow search algorithm/Dense Convolutional Network/Feature extraction/Deep learning/Image classification引用本文复制引用
出版年
2024
NETL
NSTL
万方数据