An Improved Transfer learning Approach for Intrusion Detection

Its crucial for financial systems to have sound security measures in place. For security reasons customers are not allowed to wear a helmet while using ATM(Automated Teller Machine). An automated helmet detection using ATM surveillance camera feed can help improve security significantly. Recently deep convolutional neural network (DCNN) have shown state of the art accuracy in object detection and localization. In this work, a pretrained Google's inception model have been used and have achieved an accuracy of 95.3% by training the model on a proprietary ATM surveillance dataset. Transferred information from inception model has been feed to multiple fully connected layers with drop outs to achieve better accuracy.