Badminton is a very popular subject in Physical Education (PE). Many students enroll badminton courses in every semester which pose a tremendous teaching load to the instructors. The one-on-one guiding/feedback time p...
详细信息
Coronary artery disease is the most prevalent type of heart disease and has a considerable mortality rate. In diagnosing and assessing coronary artery disease, physicians must integrate a variety of clinical informati...
详细信息
Virtualization technologies are still growing bigger and faster. Despite the greatness of its advancement, the costume industry is still very accessible when it comes to real trials. Off-the-shelf stuff are inadequate...
Virtualization technologies are still growing bigger and faster. Despite the greatness of its advancement, the costume industry is still very accessible when it comes to real trials. Off-the-shelf stuff are inadequate details for the desired individual to assess its in-depth utility for each garment trying on for a second, including custom stuff are much harder to try out right away. To this end, 2D image-based 3D reconstruction inclusive of touchable-virtualized space is accessible easier to stuff details for mans' decision making in purchasing. We establish the overall end-to-end pipeline from reconstruction until visualization for one instance to be triable on its stuff for a moment. As an expectation, our proposed approach can bring objects into the experimental area and use them immediately without any obstacle.
Scholars have long demonstrated the effectiveness of using visual representations to aid non-native speakers in learning Mandarin tone. Among these methods, presenting pitch contours visually has proven to be the most...
详细信息
In this work, we analyzed numerically a multiscale nanosystem based on sMIM on TBG. Spontaneous formation of a water-meniscus by the approximation between the tip-sample concentrates the microwave fields, reaching res...
Skin cancer is one of the most common types of cancer in the world. Different computer-aided diagnosis systems have been proposed to tackle skin lesion diagnosis, most of them based in deep convolutional neural networ...
详细信息
This study is related to a system that enables elderly people to communicate interactively with young people who use existing message exchange services by simply speaking to an avatar on a tablet PC, without having to...
详细信息
ISBN:
(纸本)9781665492331
This study is related to a system that enables elderly people to communicate interactively with young people who use existing message exchange services by simply speaking to an avatar on a tablet PC, without having to perform complicated operations of the tablet PC. On the other hand, the young people side can bring in elderly people who could not previously participate in existing message exchange services groups. The effectiveness of the system was verified through demonstration experiments with two households consisting of elderly people in their 80s.
We theoretically propose and experimentally demonstrate nonlocal metasurfaces rationally designed using symmetry-breaking principles to manipulate the optical wavefront with a geometric phase selective to custom ellip...
The existence of fine-grained image classification supporting smart retail provides effectiveness in recognizing products with high similarity. However, the generic classification method performs poorly in identifying...
详细信息
ISBN:
(数字)9798350327472
ISBN:
(纸本)9798350327489
The existence of fine-grained image classification supporting smart retail provides effectiveness in recognizing products with high similarity. However, the generic classification method performs poorly in identifying products from a subordinate category. This paper aims to identify augmentation techniques to leverage the Vision Transformer (ViT) model to classify the fine-grained grocery product, which involves embedding patches and transformer encoders to extract the main features. First, we develop a fine-grained image dataset with ColorJitter, CutOut, and combining both augmentations. Secondly, we perform experiments and analysis of ViT size, embedded patch size and image size in the patch embedding process. Lastly, the ViT model are evaluated according to the image sizes 224, 384, and 512 in accuracy, loss, and confusion matrix. The highest accuracy was obtained at 0.9922. The ColorJitter and CutOut improved the confusion matrix in ViT-B/16 and ViT-L/16 with an image size of 384 and 512. The results show that both augmentations in the ViT model are able to distinguish fine-grained grocery products.
Despite Indonesia's leading position in palm oil processing, the inaccurate assessment of oil palm fruit maturity poses challenges in determining the optimal harvest timing and maintaining product quality. With ad...
Despite Indonesia's leading position in palm oil processing, the inaccurate assessment of oil palm fruit maturity poses challenges in determining the optimal harvest timing and maintaining product quality. With advancements in information technology, including image processing and object detection techniques like YOLO and EfficientDet, there is potential for automated and precise identification of palm fruit maturity. To explore this, the authors conducted research to identify the most suitable algorithm for detecting the ripeness of palm oil fruit. The study utilized a dataset consisting of 8299 images with six maturity levels, employing augmentation techniques to increase image variability. Results show that YOLO V8s outperformed EfficientDet Lite 4 in terms of object detection precision, achieving higher mean average precision (mAP) scores. YOLO V8s showed a 4% improvement in mAP@50 and a 16% improvement in mAP@50-95 compared to EfficientDet Lite 4. YOLO V8s achieved a mAP of 0.978 for mAP@50 and 0.765 for mAP@50-95, surpassing EfficientDet Lite 4. However, EfficientDet Lite 4 had an advantage in terms of class accuracy and bounding box placement, as indicated by smaller box loss and cIs_ loss values. YOLO V8s also demonstrated better frames per second (FPS) performance, ranging from 10 to 15 FPS, compared to EfficientDet Lite 4.
暂无评论