An endoscope is a medical instrument that acquires images inside the human body. This paper proposes new approach for automatic detection of polyp regions in an image using Hessian Filter and machine learning approaches. The improves performance with higher accuracy. uses HOG feature as local since non-polyp region often have similar color information. also Real Adaboost Random Forests classifiers which works effciently even when dimension vector becomes large. It suggested filter can...

The population of elderly persons continues to grow at a high rate, and fall accidents in have become major public health problem. Highly developed IoT technology machine learning enable the use multimedia devices wide variety person's protection areas. In this paper, HOG-SVM based detection system for is proposed. To ensure privacy order be robust changes light intensity, deep sensor employed instead RGB camera get binary images persons. are detected tracked by Microsoft Kinect SDK,...

The modulation recognition of digital signals under non-cooperative conditions is one the important research contents here. With rapid development artificial intelligence technology, deep learning theory also increasingly being applied to field recognition. In this paper, a novel signal algorithm proposed, which has combined InceptionResNetV2 network with transfer adaptation, called InceptionResnetV2-TA. Firstly, received preprocessed and generated constellation diagram. Then, diagram used...

Writer recognition is very important for man-machine interface and security. Because each writer has a particular writing form, it possible to use handwriting characters recognition. But character not stable; changes time time. This makes difficult. To overcome this difficulty, we apply fuzzy theory, because can absorb the instability of by membership functions. We propose new method recognize writers in short with simple algorithm. Three types functions are obtained from normalized...

LiDAR data contain feature information such as the height and shape of ground target play an important role for land classification. The effect convolutional neural network (CNN) extraction on is very significant, however CNN cannot resolve spatial relationship features adequately. capsule (CapsNet) can identify variations widely used in supervised learning. In this article, CapsNet combined with residual (ResNet) to design a deep network-ResCapNet improving accuracy represents by vectors,...

Glomerulus segmentation in kidney tissue segments is a crucial nephropathology process used to diagnose renal diseases effectively. This study proposes novel and robust application of MLP (Multi-Layer Perceptron) based architectures for the glomeruli PAS (Periodic AcidSchiff) stained whole images effective diagnosis diseases. For challenge, proposed unique solution uses MLP-UNet Perceptron U-Net), design that evades using conventional convolution self-attention mechanisms. Additionally,...

A photometric method called point source illuminating stereo is proposed for determining the 3D shape of an object from multiple shading images under light illumination. When surface a perfect diffuser with uniform reflectance, algorithm determination positions developed by using least squares and relying on principle monocular vision inverse square law illuminance. In method, number necessary four general surface, can be reduced to three continuous surface.< <ETX...

In this paper, we propose a transfer learning based method to classify PCB images into two classes - i) True Defect and ii) Pseudo A pre-trained Inception-V3 model is used for the mid-level representations of are extracted from output an intermediate layer model. These train small adaptation network, which makes overall network suitable classification images. order tackle overfitting, regularization strategies implemented. The results our experiment, conducted on real world images, show...

This paper proposes a new approach to improve the classification accuracy of true defect and pseudo electronic circuit board. The proposed introduces detection which corresponds color image concept random sampling using multiple SVMs. first detects candidate region with high based on difference between test reference image, then extracts features recognize or defect. Data for subsets feature selection is applied find effective combination features. Selected are used recognition by each SVM...

This paper proposes a method for augmenting learning data of road damage dataset considering the influence augmented on classification accuracy. Data augmentation is very important task in field machine because more causes increasing accuracy general. The quality influences classification. Effective needed. proposed generates by selecting effective methods depending class damage. uses You Only Look Once v3 (YOLOv3) detection and an image. It tuned adding to presented public. experimental...

In this article, an automatic polyp detection system for endoscopic video frames is proposed. Manual inspection of each frame localization in the colonoscopic has many adversaries. This work proposes a real-time tracking framework region segmentation hugely acquired frames. our work, roughly detected by saliency map at first, followed modified mechanism localization. The suggests use visual as measurement model tracking. composed four probability maps generated incorporating characteristics...

In this paper, a method is proposed for colonic polyp classification which can perform virtual biopsy assessing the stage of malignancy in polyps. Geometry, texture, and colour give sufficient cue its nature. The framework characterizes geometry or shape by pyramid histogram oriented gradient (PHOG) features. To encapsulate texture surface, fractal weighted local binary pattern (FWLBP) descriptor employed, robust to affine transformation. It also partially illumination variations generally...

Semantic segmentation is one of the computer vision tasks which widely researched at present. It plays an essential role to adapt and apply for real-world use-cases, including application with autonomous driving systems. To further study self-driving cars in Thailand, we provide both proposed methods dataset this paper. In method, contribute Deeplab-V3-A1 Xception, extension DeepLab-V3&#x002B; architecture. Our method as DeepLab-V3-A1 Xception enhanced by different number <inline-formula>...

BACKGROUND: Although accurate preoperative diagnosis of lymph node metastasis is essential for optimizing treatment strategies low rectal cancer, the accuracy present diagnostic modalities has room improvement. OBJECTIVE: The study aimed to establish a high-precision method cancer using artificial intelligence. DESIGN: A retrospective observational study. SETTINGS: single center and college engineering in Japan. PATIENTS: Patients with adenocarcinoma who underwent proctectomy, bilateral...

Roadway area calculation is a novel problem in remote sensing and urban planning. This paper models this as two-step problem, roadway extraction, calculation. extraction from satellite images that has been tackled many times before. proposes method using pixel resolution to calculate the of roads covered images. The proposed approach uses U-net Resnet architectures called U-net++ ResNeXt. state-of-the-art model combined with efficient post-processing improve overlap ground truth labels....

Capsule endoscopy (CE) has been used as a reliable diagnostic measure for detecting polyps in the gastrointestinal (GI) tract. In this paper, we propose an automated framework efficient segmentation of such from endoscopic images. The is challenging task owing to differences in-context each frame. Principal componet pursuit (PCP) followed by well-known active contour (AC) model proposed method. PCP specularity removal and background subtraction our A preprocessing algorithm improve...

Light detection and ranging (LiDAR) is a frequently used technique of data acquisition it widely in diverse practical applications. In recent years, deep convolutional neural networks (CNNs) have shown their effectiveness for LiDAR-derived rasterized digital surface models (LiDAR-DSM) classification. However, many excellent CNNs too parameters due to depth complexity. Meanwhile, traditional spatial redundancy because different convolution kernels scan store information independently....

This paper proposes a new defect classification method of electronic board using Dense SIFT and CNN which can represent the effective features to gray scale image. Proposed does not use any reference image keypoints are detected on candidate region. Removing feature points except region Bag Features used histogram features. SVM judge or not. is further introduced classify true pseudo defect. Classification accuracy was evaluated effectiveness proposed shown.

Heterogeneous defect prediction (HDP) refers to using heterogeneous data collected by other projects build a model predict the software modules prone defects in project. Traditional methods usually involve measurement of source project and target However, due limitations laws regulations, these original are generally not easy obtain, which forms island. As new machine learning paradigm, federated (FL) has great advantages training In order solve island heterogeneity HDP, we propose novel...