Video Streaming Research Articles

Real-Time Skin Disease Detection and Classification Using YOLOv8 Object Detection for Healthcare Diagnosis

Skin diseases are one of the most extensive and difficult to manage topics in healthcare, affecting millions of people worldwide. Current manual diagnosis by healthcare professionals is time-consuming and dependent on an individual’s experience, authority or a medical professional’s subjective opinion, so it is paramount to create new methods to evaluate the severity of skin damage. Deep learning and computer vision technologies focused on automating the process of diagnosing and classifying skin diseases are among the most promising areas. This paper investigates the usage of webcam video streams for real-time detection and categorization of 35 different skin conditions using the YOLOv8 object detection platform. The focus will be on gathering a large and diverse dataset with annotations from Labelme and training the YOLOv8m to accurately identify and delimit regions of interest to be classified later. Key components include the setup of the Anaconda environment, installation of necessary dependencies, dataset preparation, model configuration, training procedure, and evaluation metrics. Real-time inference using webcam feed facilitates continuous monitoring and detection, providing a practical tool for dermatologists and healthcare professionals.

Act Like a Woman, Play Like a Man: Manhood Acts and the Gendered and Racialized Organization of Online Professional Streamers

Professional gaming is a billion-dollar industry with some earning incomes comparable to those of popular athletes and entertainers. The rapid rise of professional gaming owes its success to the advent of Twitch.tv, a video streaming platform that enables streamers to broadcast a live feed of their gameplay while interacting with fans. While white men have mostly dominated this arena, white women and people of color are beginning to rise in the ranks of professional streaming. In this article, we examine how online platforms like Twitch represent a new type of workplace that is organized around geek masculinity and manhood acts, establishing and perpetuating hierarchies of masculine dominance and white privilege. Analyzing interaction patterns of streamers and their viewers via publicly available text and video data, we find that men streamers and their audiences create a hostile work environment for white women and people of color online in three ways. First, gendered communication patterns of streamers uphold the gender hierarchy. Second, communication patterns of the audience rely on racialized manhood acts that put women in their “place” and perpetuate white supremacy through racialized stereotypes. Finally, manhood acts based on sexual harassment towards women, including racial epithets, signal male dominance and the dominance of white culture. These virtual manhood acts perpetuate an organizational structure of sexism and racism that establishes a hierarchical workplace, placing white men “geeks” at the top and reinforcing gender and racial inequalities in the workplace.

IMMEDIATE OBJECT SPOTTING

Object Spotting has become one of the most essential and challenging tasks in computer vision, particularly in the context of deep learning advancements. This paper reviews the evolution of object detection techniques, highlighting how deep neural networks have transformed the field. We classify the state-of-the-art detection algorithms into three main categories: anchor-based, anchor-free, and transformer-based detectors, each with distinct methodologies for identifying objects in images. This survey includes a comparison of key convolutional neural network (CNN) architectures used for object detection, evaluating their speed-accuracy trade-offs, quality metrics, and training strategies. We discuss the strengths and limitations of these models and offer insights into the future directions of research in object spotting. In parallel, immediate object spotting using OpenCV has gained prominence for its ability to identify and localize objects within live video streams, particularly by leveraging pre-trained models like YOLOv3. This approach processes each video frame in real time, with preprocessing steps such as resizing and normalization to ensure compatibility with the deep learning model. YOLOv3’s efficiency enables it to detect and classify objects quickly, drawing bounding boxes around detected objects and displaying them in real time. This project highlights how combining OpenCV with deep learning models can achieve effective, scalable object spotting while ensuring detection accuracy. The paper also explores the challenges and potential future advancements in improving detection accuracy, robustness, and compatibility with modern hardware systems for diverse applications. Keywords:Immediate object spotting, OpenCV, YOLOv3, Deep learning models, Live video stream, Pre- processing (resizing, normalization), Object identification and localization, Bounding boxes, Detection accuracy, System compatibility

Location-Based Hybrid Video Streaming Protocol for VANETs

The services and applications provided by video streaming over Vehicular Ad-hoc Networks (VANETs) have recently become very attractive to technology users. In VANETs, the moving vehicles on the roads take great advantage of this radio link by exchanging information with each other related to the safety, traffic jams, accidents, and condition of the roads via video streaming. On the other hand, this transmission domain poses many challenging issues like frequent disconnections due to fast-speed vehicles, longer transmission delays, and inefficient link consumption etc. numbers of routing strategies have been proposed in the literature that aimed to address these challenges of an Intelligent Transportation System (ITS). This article is based on a Location-Based Hybrid Video Streaming Protocol (LBHVSP) that has been proposed with the intent to minimize the challenges related to video transmission as well as the attainment of high network performance. The idea presented is a motivation from the dynamic nature of transceiver and receiver transmission power, accompanied by the range-based localization techniques for inter-space computation of vehicular nodes. LBHVSP is a 2-way protocol that works as a combination of infrastructure-based and infrastructure-less wireless networks that tends to minimize the average packet delay for efficient utilization of bandwidth. It is a two-phase solution, firstly frames are transmitted to the second hop node available in the direct range that minimizes the traveling path in terms of hop count and ultimately yields minimal delay. Secondly, this technique utilizes the Road Side Units (RSUs) having wired connectivity for a faraway destination. Experimental results using different transmission scenarios reveal its efficacy by attaining maximum average packet throughput, minimum average packet drop rate, and minimum average per-packet delay in Signal Transmission Power and Packet Reception Power Threshold values. The proposed scheme has outperformed the state-of-the-art transmission scenarios in terms of static parameters. Varying transmission and reception powers subject to the inter-distance have shown remarkable outcomes. In comparison to pre-existing VANET protocols like DyTE and DSR, proposed scheme has shown an average improvement of 30% in packet drop rate, and 80% rise in throughput, approximately. Besides, our scheme has shown a remarkable decline in transmission delay as well. On comparing the LBHVSP performance against recent VANET communication protocol, it comes out that it attains the maximum throughput and delay improvement. Furthermore, to cater the packet drop, a smart re-transmission strategy should be designed particularly for multimedia streaming and video message exchange.

Optimized Edge-Cloud System for Activity Monitoring Using Knowledge Distillation

Driven by the increasing care needs of residents in long-term care facilities, Ambient Assisted Living paradigms have become very popular, offering new solutions to alleviate this burden. This work proposes an efficient edge-cloud system for indoor activity monitoring in long-term care institutions. Action recognition from video streams is implemented via Deep Learning networks running at edge nodes. Edge Computing stands out for its power efficiency, reduction in data transmission bandwidth, and inherent protection of residents’ sensitive data. To implement Artificial Intelligence models on these resource-limited edge nodes, complex Deep Learning networks are first distilled. Knowledge distillation allows for more accurate and efficient neural networks, boosting recognition performance of the solution by up to 8% without impacting resource usage. Finally, the central server runs a Quality and Resource Management (QRM) tool that monitors hardware qualities and recognition performance. This QRM tool performs runtime resource load balancing among the local processing devices ensuring real-time operation and optimized energy consumption. Also, the QRM module conducts runtime reconfiguration switching the running neural network to optimize the use of resources at the node and to improve the overall recognition, especially for critical situations such as falls. As part of our contributions, we also release the manually curated Indoor Action Dataset.

Research on a Panoramic Image Stitching Method for Images of Corn Ears, Based on Video Streaming

Background: Corn is the main grain crop grown in China, and the ear shape index of corn is an important parameter for breeding new varieties, including ear length, diameter, row number of ears, row number of grains per ear, and so on. Objective: In order to solve the problem of limited field of view associated with computer detection of the corn ear shape index against a complex background, this paper proposes a panoramic splicing method for corn ears against a complex background, which can splice 10 corn ear panoramic images at the same time, to improve information collection efficiency, display comprehensive information, and support data analysis, so as to realize automatic corn seed examination. Methods: A summary of corn ear panoramic stitching methods under complex backgrounds is presented as follows: 1. a perceptual hash algorithm and histogram equalization were used to extract video frames; 2. the U-Net image segmentation model based on transfer learning was used to predict corn labels; 3. a mask preprocessing algorithm was designed; 4. a corn ear splicing positioning algorithm was designed; 5. an algorithm for irregular surface expansion was designed; 6. an image stitching method based on template matching was adopted to assemble the video frames. Results: The experimental results showed that the proposed corn ear panoramic stitching method could effectively solve the problems of virtual stitching, obvious stitching seams, and too-high similarity between multiple images. The success rate of stitching was as high as 100%, and the speed of single-corn-ear panoramic stitching was about 9.4 s, indicating important reference value for corn breeding and disease and insect detection. Discussions: Although the experimental results demonstrated the significant advantages of the panoramic splicing method for corn ear images proposed in this paper in terms of improving information collection efficiency and automating corn assessment, the method still faces certain challenges. Future research will focus on the following points: 1. addressing the issue of environmental interference caused by diseases, pests, and plant nutritional status on the measurement of corn ear parameters in order to enhance the stability and accuracy of the algorithm; 2. expanding the dataset for the U-Net model to include a wider range of corn ears with complex backgrounds, different growth stages, and various environmental conditions to improve the model’s segmentation recognition rate and precision. Recently, our panoramic splicing algorithm has been deployed in practical applications with satisfactory results. We plan to continue optimizing the algorithm and more broadly promote its use in fields such as corn breeding and pest and disease detection in an effort to advance the development of agricultural automation technology.

Multifeature extraction based MobileViTv3 model for fish feeding behavior recognition from video streaming

The recognition of fish feeding behavior based on machine vision is essential for optimizing fish feeding strategies and enhancing the efficiency of aquaculture. Building an efficient, multi-feature extraction model for fish feeding recognition, especially on mobile and edge devices, remains a significant challenge. In the paper, we propose a novel multi-feature extraction (MFE)-MobileViTv3 model, which improve MobileViTv3 with the MFE blocks. It can extract spatio-temporal features while obtaining the lightweight characteristic. The MFE block is designed by improving ActionNet, with the frequency channel attention (FCA) and multi-head self-attention (MHSA) mechanisms. It can fully extract spatio-temporal, motion, and channel features from video streaming, thereby further improving the feature extraction capabilities, which subsequently enhances the model's recognition accuracy. The experiments were carried out on an industrial aquaculture farm. We built a dataset of Micropterus salmoides, and then conducted the compare experiments. Compared with C3D, R3D, ResNet50, SlowFast, AlexNet, and MobileNetV3, our model can achieve a classification accuracy of 96.7 % for feeding intensity, with fewer parameters (0.96 M) and FLOPs (8.44G). The results show that the proposed model can effectively recognize fish feeding behavior with fewer parameters. Additionally, we introduce two evaluation metrics for the feeding process: Average Feeding Intensity and Strong Feeding Ratio. The metrics are conducive to the quantitative evaluation of the fish's vigor and health status.

Digital Recording and the Hazards of Unbounded Moralized Judgment

In today’s techno-social environment, it is easy to make, store, and share digital recordings, such as photographs, audio fragments, and video streams, at an unprecedented scale. While there are often obvious immediate benefits to making and sharing digital recordings, serious hazards associated with these practices have thus far gone under-appreciated. We contend that today’s digital recording practices threaten to radically alter how we perceive and evaluate ourselves and others, producing an ongoing, socially and morally disruptive shift towards unbounded moralized judgment. The shift toward unbounded moralized judgment in turn poses several hazards, including widespread, difficult-to-restore reputation damage, negatively altered self-perceptions, and the stifling of morally right behavior. Our central claim is that in the current techno-social environment, every individual has a pro tanto reason to avoid being recorded and to avoid recording others. On the occasions where the reasons for recording outweigh those against, more must be done to counteract the hazards introduced by recording. We conclude the paper by outlining possible avenues for technical, regulatory, and societal approaches to mitigating the hazards of unbounded moralized judgment.

Optimizing Video Streaming in Dynamic Networks: An Intelligent Adaptive Bitrate Solution Considering Scene Intricacy and Data Budget

Optimizing Video Streaming in Dynamic Networks: An Intelligent Adaptive Bitrate Solution Considering Scene Intricacy and Data Budget

HiVAT: Improving QoE for Hybrid Video Streaming Service With Adaptive Transcoding

HiVAT: Improving QoE for Hybrid Video Streaming Service With Adaptive Transcoding

Edge Selective Sharing for Massive Mobile Video Streaming With Cross-Layer Optimization

Edge Selective Sharing for Massive Mobile Video Streaming With Cross-Layer Optimization

Facial Feature Localization and Analysis Using Machine Learning

The facial feature localization system with machine learning (ML) is an innovative solution designed to accurately detect and locate key facial landmarks. By integrating advanced ML algorithms, the system processes images or video streams in real time to identify the positions of facial features such as the eyes, nose, mouth, and jawline. The ML component analyses the data to precisely locate these features, enabling applications in areas like facial expression analysis and biometric verification. This approach enhances accuracy by adapting to various lighting conditions, face orientations, and individual facial variations. The system’s scalability and flexibility further extend its utility, supporting diverse applications across industries like healthcare, security, and human-computer interaction. Overall, this project represents a significant advancement in facial feature localization, using IoT and ML to enable more effective and accurate facial analysis. By adapting to various lighting conditions, face orientations, and individual facial variations" to "by adapting to diverse lighting conditions, different face orientations, and individual facial variations, ensuring reliable performance in various environments.

Real-Time People Counting System

This paper introduces an advanced system for real-time people counting, utilizing live video streams or IP cameras. The core of this system is the Single Shot Multibox Detector (SSD) framework, enhanced with MobileNet architecture to ensure precise object detection. To continuously track detected individuals, we employ a centroid tracking algorithm. The system's design prioritizes scalability, making it ideal for various business applications, such as monitoring foot traffic in stores, buildings, and shopping malls. A key feature of our system is its ability to send immediate alerts when the number of people exceeds a set threshold. This functionality is crucial for managing crowd control and adhering to safety protocols, particularly in scenarios like the COVID-19 pandemic. We have optimized the system’s performance with threading, which improves its responsiveness and efficiency. By integrating advanced machine learning techniques with practical business applications, our real-time people counting system offers a dependable and scalable solution for contemporary crowd management challenges. Features like real-time alerts, automatic scheduling, and comprehensive data logging highlight its effectiveness in enhancing operational efficiency and safety across various settings.

Water Trash Detection System: An Innovative Approach to Monitoring Aquatic Pollution

Abstract: Water pollution, particularly from non-biodegradable waste such as plastics, is one of the most pressing environmental issues facing aquatic ecosystems today. This paper presents an innovative solution for detecting and classifying waste in water bodies using machine learning techniques, specifically You Only Look Once (YOLO) and Convolutional Neural Networks (CNNs). The Water Trash Detection System (WTDS) can process both static images and real-time video streams to efficiently detect various types of aquatic trash, including plastic bottles, bags, and fishing debris. The system generates detailed reports on detected waste, providing environmental organizations and researchers with actionable data for monitoring and cleaning polluted water bodies. The combination of YOLO for real-time object detection and CNNs for waste classification allows the WTDS to deliver high accuracy and real-time processing, which is essential for effective environmental management.

Exploring smartphone use and its applicability in academic training of medical students in Latin America: a multicenter cross-sectional study

BackgroundMobile phones are essential tools worldwide, including for the academic training of medical students. However, their role in Latin America (LATAM) remains underexplored. This study aimed to evaluate cell phone use and its applicability in the academic training of medical students in LATAM.MethodsA cross-sectional online survey was conducted among undergraduate medical students from seven Latin American countries between December 2023 and February 2024. The study assessed academic smartphone use across five types of applications: medical apps, knowledge updating apps, academic podcast apps, video streaming apps, and social media apps.ResultsA total of 1,590 medical students participated, with 63.6% being women, 21.2% third-year students, and 55.0% from private universities. Nearly all participants (98.4%) owned a mobile phone, and 88.2% used them for academic purposes. The most common academic use was for viewing academic content videos (82.9%). A significant association was found between academic cell phone use and progression through medical studies, with fifth-year students being the most likely to use their phones for academic purposes (OR = 4.517, 95% CI = 1.843–11.075). Students in the clinical stage of education were also more likely to rely on smart phones for academic purposes (OR = 2.090, 95% CI = 1.500–2.911). Conversely, public university students were less likely to use their phones for academic purposes compared to private university students (OR = 0.714, 95% CI = 0.522–0.977). Additionally, students who perceived that their academic preparation would be more difficult without a cell phone were more likely to use it for academic purposes (OR = 2.157, 95% CI = 1.463–3.181).ConclusionsThese results suggest the importance of incorporating smartphones into the educational frameworks of medical schools and universities in Latin America, with a particular emphasis on academic video content applications, which already capture the attention and preference of medical students.

Facial Recognition Using Video Streaming

Facial recognition has become an essential technology in numerous applications, including security, surveillance, and human-computer interaction. With the advent of real-time video streaming, facial recognition systems are now tasked with handling the continuous flow of visual data. This paper explores the integration of facial recognition algorithms into video streaming systems, investigating the challenges, methodologies, and applications of real-time face detection and identification. We present a framework for implementing facial recognition in live video streams, assess the performance of different algorithms, and highlight areas for future improvement, including handling large datasets, optimizing computation, and ensuring privacy.

SUPPRESSION OFN THE VIDEO STREAM FRAMES PROCESSED BYUNMANNED SYS-TEMS USING FPV CONTROL

Using packet networks for first-person control of unmanned systems arises a problem of large transmitted data volumes . The largest volume of traffic during first-person control is presented by video stream frames. So, to improve the efficiency of the communication network between unmanned systems and external pilot station, it is necessary to compress video stream frames. A high compression degree can be provided by using variational autoencoders. One of the problems of using variational autoencoders for frame compression is the occurrence of specific artifacts in frames. This article proposes methods for suppressing the occurrence of artifacts when restoring frames from the latent space by a neural network decoder, as well as an empirical scale for assessing autoencoder artifacts. The approach proposed encompasses preparing pixel data of a video stream frame for encoding and further reconstruction after decoding. It is experimentally shown that one of the proposed methods allows eliminating the absolute majority of artifacts without introducing significant distortions into the reconstructed frames.

Practically High Performant Neural Adaptive Video Streaming

Despite offering early promise, Deep Reinforcement Learning (DRL) suffers from several challenges in adaptive bitrate streaming stemming from the uncertainty and noise in network conditions. However, in this paper, we find that although these challenges complicate the training process, in practice, we can substantially mitigate their effects by addressing a key overlooked factor: the skewed input trace distribution in DRL training datasets. We introduce a generalized framework, Plume , to automatically identify and balance the skew using a three-stage process. First, we identify the critical features that determine the behavior of the traces. Second, we classify the traces into clusters. Finally, we prioritize the salient clusters to improve the overall performance of the controller. We implement our ideas with a novel ABR controller, Gelato , and evaluate the performance against state-of-the-art controllers in the real world for more than a year, streaming 59 stream-years of television to over 280,000 users on the live streaming platform Puffer. Gelato trained with Plume outperforms all baseline solutions and becomes the first controller on the platform to deliver statistically significant improvements in both video quality and stalling, decreasing stalls by as much as 75%.

A systematic review on WebRTC for potential applications and challenges beyond audio video streaming

AbstractVideo conferencing and live streaming are being used in various industries, such as healthcare, gaming, telecommunication, manufacturing and others. As technology progresses, the need for real-time data transmission with minimal latency has increased. Web Real-Time Communication () addresses this need effectively. WebRTC is a technology designed to provide real-time communication through web and mobile browsers. Its low latency and P2P communication capabilities make it a convenient technology for secure, efficient communication in real-time applications. This paper reviews the key features of WebRTC, discusses its strengths and weaknesses and investigates a detailed analysis of 83 existing studies. Moreover, It evaluates all use cases that can be adopted by WebRTC by examining their descriptions, problem statements, and research gaps based on literature to date. Finally, It highlights the open research directions for the emerging technologies and enhancements of WebRTC. to identify their potential applications.

Research on machine vision online monitoring system for egg production and quality in cage environment

In the domain of egg production, the application of automation technologies is essential for boosting productivity and quality. This study introduces an online monitoring system designed for egg quality assessment within caged environments, incorporating a robotic patrol system for egg localization and a fixed video stream for quality analysis. The project involved upgrading traditional henhouses with enhanced wireless connectivity and developing data transmission techniques for video streams and image data. The core of the system, an enhanced You Only Look Once Version 8-small (YOLOv8s) model, was augmented by substituting the Residual Network-18 backbone and integrating the Shuffle Attention mechanism, significantly improving egg detection precision. This refined model was implemented on Jetson AGX Orin industrial computer to facilitate real-world applications. To diverse operational needs, two distinct post-processing algorithms were developed: one for counting eggs and detecting abnormalities during robotic patrols, and another for assessing egg quality through fixed video streams, which measured crucial parameters such as egg dimensions and shape indexes. Experimental results revealed that the henhouse average network latencies of 35 ms, with signal strengths between -30 and -71 dBm, ensuring data transmission to the poultry management system. The enhanced YOLOv8s model, deployed on the Jetson AGX Orin, demonstrated well improvements: a Precision of 94.0 % (+2.4 %), Recall rate of 92.8 % (+4.6 %), Average Precision50:95 of 91.5 % (+3 %) and F1 score of 93.4 % (+3.9 %), with a minor decrease in detection speed to 91.7 Frame Per Second (-18.2). Field experiment in 60 chicken cages during robotic patrols achieved an egg recognition rate of 98.9 %, validating the system's effectiveness. In fixed settings, an 83-minute experiment managed to analyze egg numbers and abnormalities, attaining a 100 % recognition rate with all scoring data promptly relayed back to the management system. Overall, this research introduces a comprehensive system for monitoring egg production and quality in cage environments, addressing manual recording and quality assessment challenges in caged poultry farming. This study is crucial for optimizing modern livestock management, enhancing production efficiency, and ensuring animal welfare.