Deep Belief CNN Based Artificial Vision


  • Tathagat Banerjee
  • Karthikeyan S
  • Rohit Bhargav Peesa
  • Priyanka Nair


Data Analytics and Deep Learning have always tried to establish its significance to the medical scientific community. A decade ago, due to the lack of computational resources, its numerous efforts have not been able to embrace its name significantly. Today the medic community is largely astonished by the supreme pattern understanding and predictive analytics that data science under the names of deep learning and machine learning. Here in this paper we are presenting an architecture named vision. It has been used to establish a model design that can help a visually disabled people to be directed to their destination, by predicting Forward, Left and Right direction at each road step. We have maintained class division, non-overfitting and regularization at each step of our deep belief convolutional neural network which we get to know by high values for different recall and precision classes and on the other hand, the usage of Fully connected convolutional layers poses feature extraction techniques from three-dimensional images. The luxurious medical treatments are often out of reach of common public including some lethal risk factors for life, this algorithm not only solves the problem of navigation for the disabled but also path breaks this new arena of research and development. Transfer learning algorithms VGG and Inception which could attain an accuracy of 52 percent for real-time data even the results for precision and recall for low and with our architecture, we could gain accuracy of 91.41 percent along with precision and recall at an average of 89 percent.