A Fine-Tuned ConvNextSmall Approach for Accurate Footwear Classification: Model Performance and Evaluation

Emphasizing three independent categories: shoes, sandals, and boots, we study in this work the application of deep learning techniques for footwear classification. We accomplished this with a well-known performance and efficiency in image classification challenges and finely tuned ConvNextSmall convolutional neural network (CNN) architecture. The dataset employed consists of 27,000 labeled images overall; these were divided into training (12,000), validation (7,500), and testing (7,500) groups. Originally tailored to maximize performance for this specific footwear category task, the model was pre-trained using a large-scale image dataset. Several techniques of data augmentation improved generalization; the model was trained with an adaptive learning rate strategy to avoid overfitting. On the test set, evaluation with a 99% accuracy yielded quite exceptional performance for the model. This high degree of performance reveals the prospects of employing finely adjusted ConvNextSmall architectures for accurate and efficient classification in the footwear domain, therefore offering a solid solution for automated retail, inventory management, and fashion industry uses. While reducing inequality in technology access and deployment, the results enable good work and economic growth, so enabling sustainable development using responsible consumption and production, industry innovation, and solution providing.