Deep Learning Model for Fashions and Clothes Automated Classification

Abstract

In recent years, the computer vision techniques played a major role in most of the multimedia applications and one among the important criteria to classify the application is the image classification. Image classification is one of the maximum introductory glitches in computer vision. It is used extensively in most of the digital multimedia applications in association with practical applications such as video and image indexing. In spite of the major issues in identifying the multimedia images manually has set to large weaknesses for the human beings to classify the high-resolution images and it is an insignificant weakness. The generic algorithm has a lot of weaknesses in determining the accuracy. Therefore, to classify the images, the generic algorithm with an existing strategy to an invariant number of variations has been generated. Recently, a multitude of problems have been applied to deep neural networks to obtain optimal results. Specifically, convolutional deep neural networks illustrated the best results in terms of image recognition, image segmentation, problems with computer vision and issues with the representation of natural languages. The aim is to train CNN network using Sequential model classify fashion Data set image and Our task main is to create an effective model of deep learning to recognize accessories for clothing and augment Fashion-MNIST dataset that we use. The Fashion-MNIST clothing classification problem is a modern standard dataset used in computer vision and deep learning. It is relatively easy because due to sharing the exact image size, training and testing data, and format splits structure. The data has been initially pre-processed for resizing and reduce the noise. data is augmented where one image will be in three forms of output; i.e. output image is rotated, shifted and zoom as an output. Finally, the data is sent to the proposed model. The proposed model which consists of three convolutional layers that extract 32, 64, 128 filters of size 3x3 with ReLU as activation function to each layer and SoftMax in last layer , then the output of the activation function of the is fed to a max pooling of 2x2 window, then Dropout To address the problem of overfitting, finally used the ’Adam’ optimizer for optimization of the loss function. The trained model with the projected framework had 94% accuracy that achieved and compared VII with the existing works. The accuracy of the pre-Trained CNN model is used for classifying MNIST-Fashion data revealed that it is the best suited for the selected dataset.