A Hybrid Deep Learning Architecture for Image Classification Across Diverse Visual Recognition Applications

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The core computer vision problem of picture categorization has several domain-specific applications. This paper's goal is to talk about the significance of picture categorization in modern technology and society, as well as its ideas, techniques, and applications. Many computer vision systems rely on image classification, where images are automatically assigned to a predetermined category, based on the information they provide in the image. This study trains a hybrid deep learning system to efficiently recognize photos using the Caltech-256 dataset. The system makes use of both Bidirectional Long Short-Term Memory (BiLSTM) and Gated Recurrent Units (GRU). Dataset preprocessing includes standardization, label encoding, damaged image identification, and duplicate reduction. After separating the dataset into a train and test set, the following step is to extract features from each. The proposed GRU+BiLSTM model is evaluated in comparison to AlexNet, MobileNetV2, EfficientNet, and CNN models using ACC, PRE, REC, and F1-score (F1). Results for ACC (98.9%), PRE (99.1%), REC (99.3%), and F1 (100%) were better using the suggested method compared to the top deep learning models, according to the available experimental data. Findings validate the proposed hybrid design's provision of a strong and efficient