Evaluating the efficacy of deep learning models for knee osteoarthritis prediction based on Kellgren-Lawrence grading system

e-Prime - Advances in Electrical Engineering Electronics and Energy, Год журнала: 2023, Номер 5, С. 100266 - 100266

Опубликована: Авг. 31, 2023

Osteoarthritis of the knee, also known as OA has been determined that osteoarthritis knee is leading cause activity limitations and development disability, particularly in people who are older. The utilisation artificial intelligence (AI) methodologies grounded deep learning (DL) yielded promising outcomes realm radiographic interpretation. healthcare industry remarkable elevated benchmark for quality medical treatment. This study used a clinical scenario to compare twelve transfer DL models detecting grade KOA from radiograph, compared their accuracy, best model KOA. exhibited range 30% 98% It was MobileNet responsible highest level which came at 98.36%. high training validation accuracy. maximum loss observed EfficientNetB7. approaches created by skilled radiologists orthopaedic specialists could help smaller hospitals learn make more emergency room. would be especially helpful situations when personnel may not available.

Язык: Английский

---

Deep Transfer Learning Applications in Intrusion Detection Systems: A Comprehensive Review

arXiv (Cornell University), Год журнала: 2023, Номер unknown

Опубликована: Янв. 1, 2023

Globally, the external internet is increasingly being connected to industrial control systems. As a result, there an immediate need protect these networks from variety of threats. The key infrastructure activity can be protected harm using intrusion detection system (IDS), preventive mechanism that seeks recognize new kinds dangerous threats and hostile activities. This review examines most recent artificial-intelligence techniques are used create IDSs in many networks, with particular emphasis on IDS-based deep transfer learning (DTL). DTL seen as type information-fusion approach merges and/or adapts knowledge multiple domains enhance performance target task, particularly when labeled data domain scarce. Publications issued after 2015 were considered. These selected publications divided into three categories: DTL-only IDS-only works examined introduction background section, DTL-based IDS papers considered core section this review. By reading paper, researchers will able gain better grasp current state approaches different types network. Other useful information, such datasets used, employed, pre-trained network, techniques, evaluation metrics including accuracy/F-score false-alarm rate, improvements gained, also covered. algorithms methods several studies presented, principles subcategories presented reader illustrated deeply clearly

Язык: Английский

---

AI-driven behavior biometrics framework for robust human activity recognition in surveillance systems

Engineering Applications of Artificial Intelligence, Год журнала: 2023, Номер 127, С. 107218 - 107218

Опубликована: Ноя. 11, 2023

Язык: Английский

---

Improving smart home surveillance through YOLO model with transfer learning and quantization for enhanced accuracy and efficiency

PeerJ Computer Science, Год журнала: 2024, Номер 10, С. e1939 - e1939

Опубликована: Июнь 13, 2024

The use of closed-circuit television (CCTV) systems is widespread in all areas where serious safety concerns exist. Keeping an eye on things manually sounds like a time-consuming and challenging process. Identifying theft, detecting aggression, explosive risks, etc ., are circumstances which the term “security” takes multiple meanings. When applied to crowded public spaces, phrase encompasses nearly every conceivable kind abnormality. Detecting violent behaviour among them since it typically occurs group setting. Several practical limitations make hard, though complex functional difficult analyze crowd film scenes for anomalous or aberrant behaviour. This article provides broad overview field, starting with object identification moving action recognition, analysis, violence detection By combining you only look once (YOLO) transfer learning, model may acquire new skills from various sources. makes more flexible applications lessens time effort required gather large annotated datasets. proposes YOLO learning intelligent surveillance Internet Thing (IoT)-enabled home environments smart cities. Quantization concepts being optimize this work. Using quantization, optimized edge devices mobile platforms, have limited computing capabilities. Thus, even technology, be used real-world applications. proposed has been validated two different datasets 7,382 images. gains accuracy level 98.27%. method outperforms conventional one. quantization significant potential enhancing ecological city monitoring, further research development area could contribute developing effective efficient environmental monitoring systems.

Язык: Английский

---

AI-powered trustable and explainable fall detection system using transfer learning

Image and Vision Computing, Год журнала: 2024, Номер 149, С. 105164 - 105164

Опубликована: Июль 4, 2024

Язык: Английский

---