Prediction of Most Confident Labelled Output for Fashion Image using Deep Learning Model with Tensorflow

Fashion image classification is an emerging field of research. Examples of numerous e-commerce applications that rely on fashion image classification include recommendation systems, retrieval systems, and apparel recognition tasks. The extensive variety of clothing attributes, including designs, textures, and colours, and the high degree of clothing categorization present formidable challenges to the classification of clothing images. In this paper, a new algorithm called Most Confident Labelled Output (MCLO) algorithm has been proposed. Later, widely used optimization techniques in CNN models like SGD, Adam, AdaGrad and RMSProp are added for classifying the fashion images. Among the CNN optimizers, AdaGrad model achieved better accuracy as well as loss gets decreased. The proposed work was carried on benchmarking Fashion MNIST dataset The MCLO algorithm has been coded using tensorflow with graphics processing unit (GPU). The outcomes demonstrate that the proposed method outperforms state-of-the-art models, as assessed through the Fashion MNIST dataset benchmarks.