Paper Title
Foot Plantar Classification System using CNN

Abstract
The plantar of human foot being flat or cavus is common in India with high prevalence. However the proper technology is not popular as foot deformities are commonly detected using non-reliable techniques. Therefore, this paper implies to the significance of having a more scientific way to classify human foot into three classes. Even though there are many deep neural networks, we implemented Convolutional Neural Network as a tool to execute multiclass classification. A database containing hundred random human footprint images is obtained from Kaggle. In order to determine the Arch Index, the images are preprocessed into more compatible ones followed by Image Segmentation using K-Means Clustering. Subsequently the foot image is divided into three equal regions and their corresponding areas are calculated and eventually the arch index is formulated thus creating the desired dataset. Ensuingly, a CNN model is built and trained using the augmented dataset. Also, two pre-trained models, Resnet50 and MobileNet are also trained. Ultimately, a comparison is done between the custom built CNN model and the aforementioned pre-trained models for different epochs. Keywords - Economical, High Accuracy, Footprint Images, Arch Index Calculation, CNN Model.