Deep Learning-based Face Mask Usage Detection on Low Compute Resource Devices

The objective is to build an efficient face mask detector using Single Shot Detector (SSD). The algorithm used for face mask detection was a novel SSD and with the comparison of Convolutional Neural Network (CNN). The face mask detection dataset was usedand the ability of the algorithm was measured with the sample size of 136. SSD has achieved accuracy of 92.25% and for CNN it was 82.6%. By using a base architecture of VGG-16, SSD was able to outperform other object detectors like CNN without compromising speed and accuracy. The SSD and CNN are statistically satisfied with the independent sample t-test value (p<0.05) with a confidence level of 95%. Face mask detection using SSD was significantly better accurate than CNN.

Computer vision learning pay a high attention due to global pandemic COVID-19 to enhance public health service. During the fatality, tiny object detection is a more challenging task of computer vision, as it recruits the pair of classification and detection beneath of video illustration. Compared to other object detection deep neural networks demonstrated a helpful object detection with a superior achievement that is Face mask detection. However, accession with YOLOv3 covered by an exclusive topic which through certainly happening natural disease people get advantage. Added with face mask detection performed well by the YOLOv3 where it measures real time performance regarding a powerful GPU. whereas computation power with low memory YOLO darknet command sufficient for real time manner. Regarding the paper section below we have attained that people who wear face masks or not, its trained by the face mask image and non face mask image. Under the experimental conditions, real time video data that finalized over detection, localization and recognition. Experimental results that show average loss is 0.0730 after training 4000 epochs. After training 4000 epochs mAP score is 0.96. This unique approach of face mask visualization system attained noticeable output which has 96% classification and detection accuracy.

The corona virus COVID-19 pandemic is causing a global health crisis so the effective protection methods is wearing a face mask in public areas according to the World Health Organization (WHO). The COVID-19 pandemic forced governments across the world to impose lockdown to prevent virus transmissions. Reports indicate that wearing facemasks while at work clearly reduces the risk of transmission. An efficient and economic approach of using AI to create a safe environment in a manufacturing setup. A hybrid model using deep and classical machine learning for face mask detection will be presented. A face mask detection dataset consists of with mask and without mask images, we are going to use OpenCV to do real-time face detection from a live stream via our webcam. We will use the dataset to build a COVID-19 face mask detector with computer vision using Python, OpenCV, and Tensor Flow and Keras. Our goal is to identify whether the person on image/video stream is wearing a face mask or not with the help of computer vision and deep learning. As a surveillance task performer, it can also detect a face along with a mask in motion. The method attains accuracy up to 95.77% and 94.58% respectively on two different datasets. We explore optimized values of parameters using the Sequential Convolutional Neural Network of Deep Learning model to detect the presence of masks correctly without causing over-fitting.

COVID-19 pandemic has led to an international health emergency the WHO considers wearing a face mask an appropriate form of public health protection. This work will describe a face mask identification model that incorporates both deep and traditional machine learning techniques. Parts of the suggested model can be divided into two. Using Resnet50, the initial part of the system is set up for feature extraction. The second component classifies face masks using decision trees, support vector machines (SVMs), and the ensemble approach. The research will focus on three face-masked datasets. The Real-World Masked Face Dataset Includes three datasets: real-world masked faces, simulated faces, and wild faces (LFW). 99.64% of RMFD's SVM classifier is accurate throughout testing.. It achieved a 99.49% accuracy rate in SMFD and a 100% accuracy rate in LFW.

BackgroundBanana (Musa spp.) is the most popular marketable fruit crop grown all over the world, and a dominant staple food in many developing countries. Worldwide, banana production is affected by numerous diseases and pests. Novel and rapid methods for the timely detection of pests and diseases will allow to surveil and develop control measures with greater efficiency. As deep convolutional neural networks (DCNN) and transfer learning has been successfully applied in various fields, it has freshly moved in the domain of just-in-time crop disease detection. The aim of this research is to develop an AI-based banana disease and pest detection system using a DCNN to support banana farmers.ResultsLarge datasets of expert pre-screened banana disease and pest symptom/damage images were collected from various hotspots in Africa and Southern India. To build a detection model, we retrained three different convolutional neural network (CNN) architectures using a transfer learning approach. A total of six different models were developed from 18 different classes (disease by plant parts) using images collected from different parts of the banana plant. Our studies revealed ResNet50 and InceptionV2 based models performed better compared to MobileNetV1. These architectures represent the state-of-the-art results of banana diseases and pest detection with an accuracy of more than 90% in most of the models tested. These experimental results were comparable with other state-of-the-art models found in the literature. With a future view to run these detection capabilities on a mobile device, we evaluated the performance of SSD (single shot detector) MobileNetV1. Performance and validation metrics were also computed to measure the accuracy of different models in automated disease detection methods.ConclusionOur results showed that the DCNN was a robust and easily deployable strategy for digital banana disease and pest detection. Using a pre-trained disease recognition model, we were able to perform deep transfer learning (DTL) to produce a network that can make accurate predictions. This significant high success rate makes the model a useful early disease and pest detection tool, and this research could be further extended to develop a fully automated mobile app to help millions of banana farmers in developing countries.

COVID-19 has made a serious impact throughout the world. Wearing a face mask properly in public is considered an effective strategy to prevent infection. To contribute to community health, this work intends to develop an accurate real-time system for detecting non-mask faces in public. The proposed system can be used by law enforcement authorities to monitor and enforce the proper use of masks. A novel transfer learning-based method to detect face masks efficiently is proposed in this paper. Inspired by EfficientNetV2, a transfer learning model has been proposed to use for detecting whether a person is wearing a mask or not. To our knowledge, the EfficientNetV2 has not been used for detecting face masks. The model building is done based on two standard face mask datasets. The images in the first dataset consist of multiple people wearing masks, not wearing masks, and wearing the mask incorrectly. The second dataset consists of masked faces and faces without masks. To validate the generalization capability of the proposed model, the trained model is tested on two new standard datasets. In addition to that, the testing is done on a dataset created by ourselves. The proposed model performs well even on images with distortions such as blurred and noisy images. The model predicts whether the person in the input image is wearing a mask or not and also the correctness of mask-wearing with significantly better accuracy than the existing methods. The explainability of the proposed model is explained using a class activation map.

Recently the world is suffering from a pandemic. Nations were extensively affected by the Covid-19. Individuals are fighting the battle to survive; people lost their lives or the ones that matter to them while some lost their livelihood. In an hour of need everyone is trying their best to win this battle, (WHO) World Health Organization advised, to maintain at least one-meter distance, to get a vaccination, and to wear face masks. Face masks can be efficacious as a barrier between humans and the virus. The Face Mask Detection performs a critical task in suppressing the transmission of covid-19. Here, the Face masks detector was constructed with a hybrid technique. The real-time face masks detector was distributed into two segments: Training Phase and Testing Phase. The paper represents the union of the transfer learning concept, MobileNetV2 model by google, and the Caffe model, with the help of some open-source libraries used for computer vision and machine learning. The datasets were provided by the Kaggle, and they were parted into two categories: training and testing sets. The model's accuracy was compared for the two datasets. The dataset with a greater number of images has better accuracy, and to avoid overfitting data augmentation was used.

In the context of enhancing security authentication, facial recognition technology has become pivotal, replacing conventional authentication methods such as passwords, security tokens and PINs across various sectors. However, the rapid growth of facial recognition technology faced hindrances due to the COVID-19 pandemic, where mandatory face mask usage obscured facial features, challenging existing authentication methods. Regardless, the existence of several methods for face mask detection and recognition highlighted prevalent issues such as poor lighting, varied angles, failed detection for improper use of face masks, computational complexity, difficulty in detecting smaller faces and low-resolution targets have led to suboptimal accuracy rates. Hence, this study aims to address these challenges by introducing a hybrid Convolutional Neural Network (CNN) architecture tailored for Face Mask Detection (FMD) and Masked Facial Recognition (MFR). The models used MobileNetV2 and FaceNet InceptionResNetV1 respectively for FMD and MFR. The proposed models leverage advanced FMD and MFR technologies, contributing to the real-world need for enhanced security in scenarios where traditional methods are insufficient. The models underwent training using five distinct datasets, comprising a total of 10,980 images for FMD across two datasets, and 26,523 images for MFR across three datasets. In the FMD phase, the model achieved exceptional results, attaining a perfect 100% across evaluation metrics such as accuracy, precision, recall, and the f1-score within a training timeframe of an hour. Transitioning to the MFR phase, where the model required approximately one hour and 30 minutes, maintained an outcome of 99.68% across the aforementioned metrics, surpassing the accuracy level of existing models within the meta-analysis. Furthermore, the model underwent testing on a real-time custom dataset designed for MFR evaluation, consisting of 5500 images (i.e., 4400 for training, 550 for validation and 550 for testing) in real-life scenarios. Robustness was assessed under various conditions, resulting in an impressive 99.82% accuracy. The model demonstrated high accuracy in real-time testing. Notably, both the models excel in detecting and recognising masked participants from diverse angles and lighting conditions with minimal computational complexity. Leveraging the pre-trained MobileNetV2 for FMD and FaceNet InceptionResNetV1 with CNN for MFR, the CNN models provide a comprehensive solution. The proposed models surpass existing methods, excelling in accuracy under challenging conditions. This study contributes a versatile and efficient solution, addressing limitations in current approaches and providing robust models for FMD and MFR in diverse sectors.

The rapid growth of facial recognition technology has faced hindrances due to the COVID‐19 pandemic, where mandatory face mask usage obscured facial features, challenging existing authentication methods. Despite the rapid development of several methods for face mask detection and recognition that highlighted prevalent issues such as poor lighting, varied angles, failed detection for improper use of face masks, computational complexity, difficulty in detecting smaller faces and low‐resolution targets, these have led to suboptimal accuracy rates. Hence, this work addresses these challenges by introducing a hybrid convolutional neural network (CNN) architecture tailored for face mask detection (FMD) and masked facial recognition (MFR). The models are developed using MobileNetV2 and FaceNet InceptionResNetV1 with CNN, for FMD and MFR, respectively. Experimental results on both models utilising a total of five distinct datasets, with two for FMD and three for MFR, show the superiority of the developed model in comparison to state‐of‐the art models. In addition, the models are tested in real‐time for both FMD and MFR to determine their robustness, efficiency and accuracy in a real‐time context. For this purpose, a ‘custom real‐time masked face recognition’ (CRMFR) dataset was developed to perform real‐time MFR. Leveraging advanced FMD and MFR technologies, the models contribute to the real‐world need for enhanced security in scenarios where traditional methods are insufficient.

Amid current pandemic, Covid-19 has made us realize the importance of Face Masks and we need to understand the crucial effects of not wearing one, now more than ever. Right now, there are no face mask detectors installed at the crowded places. But we believe that it is of utmost importance that at transportation junctions, densely populated residential area, markets, educational institutions and healthcare areas, it is now very important to set up face mask detectors to ensure the safety of the public. In this paper we have tried to build a two phased face mask detector which will be easy to deploy at the mentioned outlets. With the help of Computer Vision, it is now possible to detect and implement this on large scale. We have used CNN/ MobileNet V2 architecture for the implementation of our model. The implementation is done in Python, and the python script implementation will train our face mask detector on our selected dataset using TensorFlow and Keras. We have added more robust features and trained our model on various variations, we made sure to have large varied and augmented dataset so that the model is able to clearly identify and detection the face masks in real time videos. The trained model was tested on both real-time videos and static pictures and in both the cases the accuracy was more than the other designed models.

Many lives have been put in danger since the start of the COVID-19 outbreak. According to WHO (World Health Organization)’s statements, breathing without a mask is highly dangerous in public and crowded places. Wearing a mask is essential as long as the vaccination is not widely used and does not fully protect everyone. Wearing masks does minimize the chances of becoming infected and the transmission of Coronavirus. Due to this, many public service providers may mandate customers to wear masks in order to get service.However, manually monitoring whether people are wearing a mask or not in public is inefficient and difficult. This paper proposes replacing manual inspection with a deep learning-based method using YOLOX, the most powerful objection detection algorithm. The results of the experiments show that the algorithm described in this work can efficiently detect face masks and enable more effective personnel monitoring than other YOLO series models. In addition, as compared to other models, our technique has significant benefits in terms of speed and accuracy in small and crowded areas.

Skin diseases in cats often present similar visual symptoms across different conditions, making early and accurate diagnosis challenging for pet owners and veterinarians. This study develops a classification model for cat skin diseases: Fungal Infection, Flea Infestation, Scabies, and Healthy, using the MobileNetV2 architecture with a transfer learning approach. A total of 1,600 RGB images were collected from public datasets and divided into 1,280 training and 320 validation samples. The dataset underwent preprocessing, normalization, and data augmentation techniques such as rotation, shear, zoom, and flipping to enhance model generalization and reduce overfitting. Several experiments were conducted to analyze the impact of input size and learning rate adjustments on model performance. The optimal configuration was achieved using an input size of 224×224 pixels, a learning rate of 0.001, and augmentation applied to the training data. The resulting model achieved a validation accuracy of 91.8%, with an average precision, recall, and F1-score of 91%, demonstrating balanced performance across all classes. These results indicate that the MobileNetV2 architecture, combined with appropriate hyperparameter tuning and augmentation, provides a reliable and computationally efficient method for automatic identification of cat skin diseases. This approach can support early diagnosis, improve animal welfare, and serve as a foundation for the development of practical veterinary diagnostic applications.

The problem faced in identifying the quality of black soybeans is that the quality of the assessment is inconsistent and it takes a relatively long time. This study aims to determine the best convolutional neural network architecture by comparing the performance of Custom CNN, MobileNetV2, and ResNet-34 architectures in identifying the quality level (grade) of black soybeans. The quality of black soybean is split into 4 different classes based on physical characteristics (split, damaged, other colors, wrinkles, dirt) and moisture content test. The number of images used is 1300 images, with the ratio of training data, validation data, and testing data are 50:25:25, 60:25:15, and 70:20:10. The best model for identifying the quality based on the physical characteristics is the MobileNetV2 architecture with a ratio of 50:25:25 which produces an accuracy of 90.18%. Morover, the best model for identifying the quality based on the moisture content is the ResNet-34 architecture with a ratio of 70:20:10, which produces an accuracy of 78.12%. The best overall accuracy in identifying the quality based on both physical characteristics and moisture content is the ResNet-34 architecture, with a ratio of 70:20:10, with an average accuracy of testing data of 79.21%.

Object Detection Technology has been a subject to much research and development due to increasing use of images and videos as data sources and their huge number of applications. Traditional models for object detection had limitations in training and did not use transfer learning for their benefit. With the evolution of deep learning and Neural networks, newer and powerful tools have made way to achieve Object Detection in real-time, with the added advantage of transfer learning and detection of multiple instances of different classes of interest in the given image context. The proposed system is an Object Detection model based on the Single Shot Detector (SSD) algorithm trained with MobileNetV2 feature extraction that can be utilized and integrated in e-commerce, hospitality industry, security and surveillance, real estate, self-driving cars and floor inventory management.

Deep Learning applied to Road Accident Detection with Transfer Learning and Synthetic Images