Scholarlink

<p>Deep learning techniques excel in balanced image classification but often struggle with imbalanced datasets, leading to poor recognition of minority-class samples. To address this, we propose a novel Convolutional Neural Network (CNN) architecture designed for imbalanced object classification. Our approach integrates an adaptive feature extraction mechanism that selectively fuses discriminative features when classification accuracy falls below a predefined threshold. Principal Component Analysis (PCA) reduces the dimensionality of feature vectors, minimizing computational complexity and memory requirements. Experimental results demonstrate that our method achieves 98.91% accuracy on the Caltech-101 dataset and 98.00% on the Pascal VOC 2012 dataset, outperforming several state-of-the-art techniques. These results validate the effectiveness of our approach in addressing class imbalance challenges.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Imbalanced object classification using a new convolutional neural network and principal component analysis

  • 期刊论文

  • 发表日期：

    2026年2月3日

查看全文
  • Homayoun Rastegar,
  • Hassan Khotanlou

摘要

Deep learning techniques excel in balanced image classification but often struggle with imbalanced datasets, leading to poor recognition of minority-class samples. To address this, we propose a novel Convolutional Neural Network (CNN) architecture designed for imbalanced object classification. Our approach integrates an adaptive feature extraction mechanism that selectively fuses discriminative features when classification accuracy falls below a predefined threshold. Principal Component Analysis (PCA) reduces the dimensionality of feature vectors, minimizing computational complexity and memory requirements. Experimental results demonstrate that our method achieves 98.91% accuracy on the Caltech-101 dataset and 98.00% on the Pascal VOC 2012 dataset, outperforming several state-of-the-art techniques. These results validate the effectiveness of our approach in addressing class imbalance challenges.