Future Generation Computer Systems, Год журнала: 2023, Номер 152, С. 343 - 345
Опубликована: Окт. 27, 2023
Advances in geospatial technologies book series, Год журнала: 2024, Номер unknown, С. 88 - 112
Опубликована: Июнь 7, 2024
Modern computer vision and machine learning technologies have enabled numerous advances in a variety of domains, including pattern recognition image classification. One the most powerful methods is capsule network, which encodes features based on their hierarchical relationships. A network sort neural that uses inverted graphics to represent an item distinct sections see existing link between these pieces, as opposed CNNs, lose evidence relating spatial placement require large amount training data. As result, authors give comparison various designs utilized diverse applications. The fundamental contribution this study it summarizes discusses major current published topologies, advantages, limits, modifications,
Язык: Английский
Процитировано
0
Advances in geospatial technologies book series, Год журнала: 2024, Номер unknown, С. 151 - 172
Опубликована: Июнь 7, 2024
The human face can appear different depending on the circumstances because of its flexibility and three-dimensional structure. Researchers are facing several obstacles relating to poses, illumination, facial expressions, head direction, occlusion, hairdo, etc. in process developing dependable efficient algorithms for detection, identification, expression analysis. To determine algorithms' effectiveness, they need be evaluated against a certain set image/database benchmarks. This work introduces dataset multiple-pose photographs. Eight hundred fifty photos from 50 people 17 distinct stances included collection (0°, 5°, 10°, 15°, 20°, 25°, 30°, 35°, 55°, -5°, -10°, -15°, -20°, -25°, -30°, -35°, -55°). Three lighting conditions also dataset. resolutions (144 × 256, 200 200, 100 100, 70 70, 50, 40 40, 20 10 pixels) available dataset's image content provide insight effectiveness resilience upcoming detection recognition systems. Additionally, based suggested database, comparison study two methods, such as PAL PCA, is performed.
Язык: Английский
Процитировано
0
Опубликована: Июнь 11, 2024
Язык: Английский
Процитировано
0
Опубликована: Окт. 9, 2024
Wireless capsule endoscopy (WCE) is a minimally invasive medical imaging technique that provides real-time visual information about the digestive system, enabling detection and diagnosis of various gastrointestinal disorders. It offers patient-friendly alternative to traditional endoscopic procedures, providing valuable insights into system's health with minimal discomfort procedures. However, accurate classification bleeding normal WCE images challenging due varying lighting conditions, artifacts, presence regions. In this study, an ensemble feature extraction approach utilizing ResNet 50, VGG 16, Inception V3 neural networks was proposed for images. The framework incorporates image processing, augmentation, data preprocessing, extraction, thereby enhancing accuracy classification. Furthermore, autoencoder employed reconstruct extracted features, Support Vector Machine classifier integrated differentiate classify experimental results demonstrated impressive average 99% precision 99.5%, showcasing efficacy method. This study contributes improving timely treatment planning disorders using imaging.
Язык: Английский
Процитировано
0
International Journal of Electrical and Electronics Research, Год журнала: 2023, Номер 11(2), С. 575 - 581
Опубликована: Июнь 30, 2023
Although gastric cancer is a prevalent disease worldwide, accurate diagnosis and treatment of this condition depend on the ability to detect lymph nodes. Recently, use Deep learning (DL) techniques combined with CT imaging has led development new tools that can improve detection disease. In study, we will focus CNNs, specifically those built “MobileNet” “AlexNet” platforms, The study begins an overview discusses importance detecting nodes in management cycle. DL are discussed as potential technologies accuracy detection. look into performance namely images patients cancer. utilizes dataset consisting individuals who have annotated Various preprocessing steps, such segmentation image normalization, carried out relevance quality data. two CNN architectures, “AlexNet”, evaluated for their area. Transfer methods utilized fine-tune models results experiments analyzed determine models' performance. findings show model more than other platforms when it comes highlights advantages using enhance suffering from It supports notion could help outcomes
Язык: Английский
Процитировано
0
Future Generation Computer Systems, Год журнала: 2023, Номер 152, С. 343 - 345
Опубликована: Окт. 27, 2023
Процитировано
0