Advancing Image Classification and Detection Using CNNs: High Accuracy, Scalability, and Generalization

Abstract

The exponential growth of visual data has spurred significant advancements in image classification and detection, with artificial intelligence (AI) playing a pivotal role in achieving high accuracy and scalability. This study presents a CNN-based methodology for image classification and object detection, leveraging supervised learning to classify over 2,000 images of cats and dogs and detect objects within augmented datasets to enhance diversity. The proposed model achieved notable results, with accuracy metrics of 91.32% for cat classification and 90.17% for dog classification, alongside high detection precision. Precision and recall metrics further underscored the model’s effectiveness, yielding F1 scores of 91.31% and 89.93% for cat and dog classifications, respectively. Object detection capabilities were evaluated using models such as YOLO and Faster R-CNN, achieving a mean average precision (mAP) of 89.2% across diverse image datasets. Comparative analyses revealed that the CNN-based approach significantly outperformed traditional regression models, with improvements exceeding 25% in classification accuracy and robust object localization performance. The model’s scalability was demonstrated through consistent results across augmented datasets, achieving an average accuracy of 90.58%. Error analysis highlighted challenges in images with overlapping features, suggesting avenues for improvement through attention mechanisms. This study provides a comprehensive framework for advancing image classification and detection systems across diverse applications, such as medical imaging, autonomous driving, and e-commerce platforms.