Classification of Gastric Lesions Using Gabor Block Local Binary Patterns

The identification of cancer tissues in Gastroenterology imaging poses novel challenges to the computer vision community in designing generic decision support systems.This generic nature demands the image descriptors to be invariant to illumination gradients,scaling,homogeneous illumination,and rotation.In this article,we devise a novel feature extraction methodology,which explores the effectiveness of Gabor filters coupled with Block Local Binary Patterns in designing such descriptors.We effectively exploit the illumination invariance properties of Block Local Binary Patterns and the inherent capability of convolutional neural networks to construct novel rotation,scale and illumination invariant features.The invariance characteristics of the proposed Gabor Block Local Binary Patterns(GBLBP)are demonstrated using a publicly available texture dataset.We use the proposed feature extraction methodology to extract texture features from Chromoendoscopy(CH)images for the classification of cancer lesions.The proposed feature set is later used in conjuncture with convolutional neural networks to classify the CH images.The proposed convolutional neural network is a shallow network comprising of fewer parameters in contrast to other state-of-the-art networks exhibiting millions of parameters required for effective training.The obtained results reveal that the proposed GBLBP performs favorably to several other state-of-the-art methods including both hand crafted and convolutional neural networks-based features.

A Hybrid Duo-Deep Learning and Best Features Based Framework for Action Recognition

Human Action Recognition(HAR)is a current research topic in the field of computer vision that is based on an important application known as video surveillance.Researchers in computer vision have introduced various intelligent methods based on deep learning and machine learning,but they still face many challenges such as similarity in various actions and redundant features.We proposed a framework for accurate human action recognition(HAR)based on deep learning and an improved features optimization algorithm in this paper.From deep learning feature extraction to feature classification,the proposed framework includes several critical steps.Before training fine-tuned deep learning models–MobileNet-V2 and Darknet53–the original video frames are normalized.For feature extraction,pre-trained deep models are used,which are fused using the canonical correlation approach.Following that,an improved particle swarm optimization(IPSO)-based algorithm is used to select the best features.Following that,the selected features were used to classify actions using various classifiers.The experimental process was performed on six publicly available datasets such as KTH,UT-Interaction,UCF Sports,Hollywood,IXMAS,and UCF YouTube,which attained an accuracy of 98.3%,98.9%,99.8%,99.6%,98.6%,and 100%,respectively.In comparison with existing techniques,it is observed that the proposed framework achieved improved accuracy.

ROBO-SPOT:Detecting Robocalls by Understanding User Engagement and Connectivity Graph

Robo or unsolicited calls have become a persistent issue in telecommunication networks,posing significant challenges to individuals,businesses,and regulatory authorities.These calls not only trick users into disclosing their private and financial information,but also affect their productivity through unwanted phone ringing.A proactive approach to identify and block such unsolicited calls is essential to protect users and service providers from potential harm.Therein,this paper proposes a solution to identify robo-callers in the telephony network utilising a set of novel features to evaluate the trustworthiness of callers in a network.The trust score of the callers is then used along with machine learning models to classify them as legitimate or robo-caller.We use a large anonymized dataset(call detailed records)from a large telecommunication provider containing more than 1 billion records collected over 10 days.We have conducted extensive evaluation demonstrating that the proposed approach achieves high accuracy and detection rate whilst minimizing the error rate.Specifically,the proposed features when used collectively achieve a true-positive rate of around 97%with a false-positive rate of less than 0.01%.