Fine-Tuned CNN for Clothing Image Classification on Mobile Edge Computing

This paper explores the use of convolutional neural networks for clothing image classification based on evaluation metrics such as F1-score, accuracy, precision, and recall. The goal is to improve the accuracy of image detection and classification to help individuals choose the appropriate clothing, recycle clothing, and help retailers supervise their inventory and optimize the goods. A comparison between four state-of-the-art models was performed, emphasizing some thresholds for these models (recall greater than 74%, precision above 89%). The ZF NET model has the highest accuracy (over 81%), which shows its effectiveness in identifying and categorizing clothing items. The VGG-16 model has the best F1-score (86.4%), which reflects a good balance between precision and recall. We also propose a fine-tuned convolutional neural network called Clothes-DAT, and compare it with the four reference models, in terms of recall, accuracy, and F1-score. Additionally, Clothes-DAT has strong generalization ability, being able to handle not only the training data sets but also new data.