Enhancing Image Classification Accuracy with Advanced Convolutional Neural Network Architectures
Mohd Sameer, Geeta Rani, Sushant Kumar, Tushar Anand, Anmol Thakur · International Journal of Research Publication and Reviews · 2025
Image classification has remained one of the in-dispensable activities in computer vision, since its applications range over a wide horizon that is inclusive of, but not limited to, medical imaging and autonomous systems.This paper surveys advanced CNN architecture that fosters higher accuracy in the image classification task.We will go into depth with several state-of-the-art CNN models including ResNet, DenseNet, and EfficientNet, which can solve major challenges like overfitting, computation complexity, and feature extraction.Among these, the performance comparison was done through extensive experi-mentation on benchmark datasets such as CIFAR-10, ImageNet, and MNIST.Our results show significant improvements in the classification accuracies due to deeper networks, residual connections, and efficient use of parameters.Further, transfer learning and data augmentation techniques have been tried to further optimize the performance of models.It therefore provides insight into the selection and utilization that could be made of the most advanced CNN architectures, which can be used in a number of various tasks of image classification, reinforcing the development of image recognition systems with enhanced accuracy and efficiency.