Enhancing Fine-Grained Image Classification u...

Enhancing Fine-Grained Image Classification using Hybrid Transformer-CNN Ensemble Models

Author Name : Basava Venkatasaivaraprasad, Pranav Vamsi Krishna Chitirala, AadiBhanuprakash, Jayanth Kumar Macha

DOI: https://doi.org/10.56025/IJARESM.2023.119231319

ABSTRACT Fine-grained image classification, the nuanced task of discerning visually similar categories, remains a persistent and demanding problem in the realm of computer vision. In response, this paper introduces an innovative approach that harnesses the synergy of hybrid models, seamlessly integrating transformer and convolutional neural network (CNN) architectures within a meticulously designed ensemble framework. While traditional CNNs have demonstrated prowess in feature extraction, their limitations become evident when confronted with the intricate visual disparities inherent to fine-grained categorization. In contrast, transformers, originally tailored for natural language processing, present an enticing avenue due to their unique ability to capture intricate contextual relationships, making them a natural fit for image analysis. Our hybrid model unfolds as a sophisticated amalgamation of architectural strengths