The Prediction of 60,000 Fashion Items by the Sequential Model in Deep Learning

Abstract

The ultimate goal of this paper is to make the sequential model predict what 60,000 pictures about fashion items are in deep learning. A point to note is that when epoch was 24, the accuracy rate of train data was highest, while when epoch was 23, the accuracy rate of validation data was highest. It is worthwhile pointing out that the accuracy of the sequential model is higher than the val_accuracy of the sequential model. This in turn indicates that the sequential model worked well for train data rather than validation data. A further point to note is that the sequential model correctly predicted that 677 fashion items were T-shirts. However, it wrongly predicted that 218 shirts were T-shirts. Quite interestingly, it wrongly predicted that 66 dresses were T-shirts. It is worth noting that there were slight fluctuations in the accuracy rate of the sequential model, but there was a gradual increase in the accuracy of the sequential model. It must be noted that the figure reached a peak when learning took place 25 times (91.20%). A major point of this paper is that the accuracy of the sequential model is slightly higher than the val_accuracy of the sequential model. This in turn suggests that the sequential model worked for train data rather than validation data. It is interesting to observe that the sequential model correctly predicted that 677 fashion items were T-shirts. However, it wrongly predicted that 66 dresses were T-shirts. Also, it wrongly judged 218 shirts as T-shirts. Finally, it must be stressed that the accuracy rate of the sequential model in test data was 83%.