Bio: Ashim Dey is an academic researcher from Chittagong University of Engineering & Technology. The author has contributed to research in topics: Facial recognition system & Convolutional neural network. The author has an hindex of 1, co-authored 8 publications receiving 2 citations.
05 Apr 2021
TL;DR: In this paper, the performance of various machine learning algorithms for predicting breast cancer was evaluated and the accuracy of each algorithm was calculated and compared to find the most suitable one, based on the analysis, Random Forest and Support Vector Machine outperformed other classifiers with accuracy of 96.5%.
Abstract: At the moment, the most prevalent form of cancer diagnosed in women across the globe is breast cancer. It develops in the breast tissue and is one of the most frequent causes of women’s death. This cancer can be cured if it is diagnosed at preliminary stage. Malignant and benign are two types of tumor found in case of breast cancer. Malignant tumors are deadly as their rate of growth is much higher than benign tumors. So, early identification of tumor type is pivotal for the appropriate treatment of a patient having breast cancer. In this work, Wisconsin Breast Cancer Dataset has been used which was collected from UCI repository. Our goal is to analyze the dataset and evaluate the performance of various machine learning algorithms for predicting breast cancer. Here, Support Vector Machine, Logistic Regression, K-Nearest Neighbors, Decision Tree, Naive Bayes and Random Forest classifiers have been implemented for classifying tumors into benign and malignant. The accuracy of each algorithm is calculated and compared to find the most suitable one. Based on the analysis, Random Forest and Support Vector Machine outperform other classifiers with accuracy of 96.5%. These classifiers can be used to build an automatic diagnostic system for preliminary diagnosis of breast cancer.
••21 Dec 2020
TL;DR: In this paper, a face recognition based automatic student attendance system using Convolutional Neural Networks is presented. But, the system can detect and recognize multiple person's face from video stream and automatically record daily attendance and achieved an average recognition accuracy of about 92 %.
Abstract: We are living in the 21st century which is the era of modern technology. Many traditional problems are being solved using new innovative technologies. Taking attendance daily is an indispensable part of educational institutions as well as offices. It is both exhausting and time-consuming if done manually. Biometric attendance systems through voice, iris, and fingerprint recognition require complex and expensive hardware support. An auto attendance system using face recognition, which is another biometric trait, can resolve all these problems. This paper represents the development of a face recognition based automatic student attendance system using Convolutional Neural Networks which includes data entry, dataset training, face recognition and attendance entry. The system can detect and recognize multiple person's face from video stream and automatically record daily attendance. The proposed system achieved an average recognition accuracy of about 92 %. Using this system, daily attendance can be recorded effortlessly avoiding the risk of human error.
••11 Dec 2020
TL;DR: A lightweight deep Convolution neural network model for real-time age and gender prediction and different experimental investigations on the prepared dataset show that with most recent approaches, this model provides competitive prediction accuracy.
Abstract: Recognition of age and gender has become a significant part of the biometric system, protection, and treatment It is widely used for people to access age-related content It is used by social media in the distribution of layered advertising and promotions to expand its scope Application of face detection has grown to a great extent that we should upgrade it using various methods to achieve more accurate results In this paper, we have developed a lightweight deep Convolution neural network model for real-time age and gender prediction For making the training dataset more diverse, Wiki, UTKFace, and Adience datasets have been merged into one containing 18728 images Using this vast mixed dataset, we have achieved accuracy of 4859% and 8076% for age and gender respectively Further, the model is tested in real-time Different experimental investigations on the prepared dataset show that with most recent approaches, our model provides competitive prediction accuracy
06 Apr 2021
TL;DR: In this article, Haar Cascade Classifier is used to detect face from a live video stream and then Local Binary Pattern Histogram (LBPH) algorithm is applied to create the recognizer for face recognition using OpenCV-Python library.
Abstract: A large number of people around the world are suffering from visual impairment which is a global health issue. These visually challenged people face a great deal of difficulties in carrying out their day-to-day activities. Recognizing a person is one of the major problems faced by them. This document represents a face recognition system with auditory output which can be beneficial for visually challenged people in recognizing known and unknown persons. Proposed face recognition system is comprised of three main modules including dataset creation, dataset training, and face recognition. Here, Haar Cascade Classifier is used to detect face from a live video stream and then Local Binary Pattern Histogram (LBPH) algorithm is applied to create the recognizer for face recognition using OpenCV-Python library. This system can detect and recognize multiple people and is also capable of recognizing from both front and side face. The overall face recognition accuracy is about 93%. Apart from visually challenged people, old people with Alzheimer’s disease can also be benefited using this system.
05 Jan 2021
TL;DR: In this paper, an automated system for identifying the Bangladeshi banknotes using a convolutional neural network has been proposed for helping visually impaired people, which is invariant of the orientation and sides of notes.
Abstract: Visually impaired people face extreme difficulties in recognizing paper currencies for day-to-day transactions as the texture and shape of many currencies are very similar. In this paper, an automated system for identifying the Bangladeshi banknotes using a convolutional neural network has been proposed for helping visually impaired people. A new dataset has been created consisting of more than 70,000 images of currently available Bangladeshi banknotes. The development of the system includes dataset building, creating and training the convolutional neural network model, and testing it in real-time. For verifying the efficacy of the proposed method, it has been tested in various backgrounds. The system can identify the eight banknotes used in Bangladesh with an average accuracy of 92% and exhibit the result with both textual and auditory output. Moreover, the system is invariant of the orientation and sides of notes. Visually impaired people will be able to easily use it in daily transactions.
••01 Jan 2021
TL;DR: In this paper, the authors presented an IoT-enabled automated object recognition system that simplifies the mobility problems of the visually impaired in indoor and outdoor environments The overall accuracy of the proposed system in object detection and recognition is 9931% and 9843% respectively.
Abstract: Background Visual impairments have become one of the most predominant problems for the last few decades To keep doing their daily tasks, vision-impaired people usually seek help from others An automated common object and currency recognition system can improve the safe movement and transaction activity of visually impaired people Objective To develop a system that can identify indoor and outdoor objects, notify the users, and send all information to a remote server repeatedly at a fixed time interval Methods The proposed system assists the visually impaired to recognize several objects and provides an audio message to aware the user Four laser sensors are used in the system to detect the objects in the direction of the front, left, right and ground The proposed system uses Single Shot Detector (SSD) model with MobileNet and Tensorflow-lite to recognize objects along with the currency note in the real-time scenario in both indoor and outdoor environments Results Among 375 participants, 82% reacted that the price of the proposed system is reasonable, 13% treated as the cost is moderate and the rest 5% people responded that the cost is relatively high for them In terms of size and weight, 73% reacted that the size and weight are considerable, 20% treated that the size is not suitable, and weight needs to lessen, and the rest 7% people responded that the system is bulky Regarding input signal observation, 98% responded that they have heard the sound appropriately and the remaining 2% of individuals missed hearing the signal Conclusions This paper represents an IoT-enabled automated object recognition system that simplifies the mobility problems of the visually impaired in indoor and outdoor environments The overall accuracy of the proposed system in object detection and recognition is 9931% and 9843% respectively In addition, the proposed system sends all processed data to a remote server through IoT
TL;DR: In this paper , an ensemble of two CNN architectures integrated with Channel and Spatial attention was proposed for breast cancer classification, which achieved a classification accuracy of 99.55% on the BreakHis dataset.
TL;DR: In this article , a systematic mapping review is mainly focused on the scene understanding aspect (e.g., object recognition and obstacle detection) of assistive solutions, and an overview of the current challenges and a comparison between different solutions is provided to indicate the pros and cons of existing approaches.
Abstract: Abstract During the past years, the development of assistive technologies for visually impaired (VI)/blind people has helped address various challenges in their lives by providing services such as obstacle detection, indoor/outdoor navigation, scene description, text reading, facial recognition and so on. This systematic mapping review is mainly focused on the scene understanding aspect (e.g., object recognition and obstacle detection) of assistive solutions. It provides guidance for researchers in this field to understand the advances during the last four and a half years. This is because deep learning techniques together with computer vision have become more powerful and accurate than ever in tasks like object detection. These advancements can bring a radical change in the development of high-quality assistive technologies for VI/blind users. Additionally, an overview of the current challenges and a comparison between different solutions is provided to indicate the pros and cons of existing approaches.
Abstract: Breast cancer is one of the most dangerous cancers, accounting for a large number of fatalities each year. It is the leading cause of mortality among women globally. It is getting a lot of interest in the scientific community because of its possible life-threatening danger. As a consequence, many machine learning methods (MLMs) have been modified to provide the best results for early diagnosis of this malignancy. Machine learning methods (MLMs) offer several beneficial implications in breast cancer, including early prognosis, detection, and diagnosis. Compared to traditional statistical analysis, machine learning methods (MLMs) have the capacity to improve the analysis of various health data, such as unstructured, complicated, and noisy data. With the demanding prevalence of breast cancer and the arrival of “data reformation,” it is thus imperative to mention the ethical consequences of machine learning (ML) on society and cancer care. It offers conclusively strong tools, smart methods, and efficient algorithms that can help in the prognosis of breast cancer. The focus of this review is on supervised techniques such as classification and regression that may be implemented and used for breast cancer data analysis. Some supervised learning methods like Naive Bayes, AdaBoost, and support vector machine are presented in this work in the early identification of breast cancer. These algorithms have been analyzed for their accuracy and efficiency using various assessment metrics and methods.
TL;DR: A Deep Learning approach for people with visual impairment that addresses the aforementioned issue with a voice-based form to represent and illustrate images embedded in printed texts with the highest prediction accuracy of an image caption.
Abstract: Recent advances in machine and deep learning algorithms and enhanced computational capabilities have revolutionized healthcare and medicine. Nowadays, research on assistive technology has benefited from such advances in creating visual substitution for visual impairment. Several obstacles exist for people with visual impairment in reading printed text which is normally substituted with a pattern-based display known as Braille. Over the past decade, more wearable and embedded assistive devices and solutions were created for people with visual impairment to facilitate the reading of texts. However, assistive tools for comprehending the embedded meaning in images or objects are still limited. In this paper, we present a Deep Learning approach for people with visual impairment that addresses the aforementioned issue with a voice-based form to represent and illustrate images embedded in printed texts. The proposed system is divided into three phases: collecting input images, extracting features for training the deep learning model, and evaluating performance. The proposed approach leverages deep learning algorithms; namely, Convolutional Neural Network (CNN), Long Short Term Memory (LSTM), for extracting salient features, captioning images, and converting written text to speech. The Convolution Neural Network (CNN) is implemented for detecting features from the printed image and its associated caption. The Long Short-Term Memory (LSTM) network is used as a captioning tool to describe the detected text from images. The identified captions and detected text is converted into voice message to the user via Text-To-Speech API. The proposed CNN-LSTM model is investigated using various network architectures, namely, GoogleNet, AlexNet, ResNet, SqueezeNet, and VGG16. The empirical results conclude that the CNN-LSTM based training model with ResNet architecture achieved the highest prediction accuracy of an image caption of 83%.