Attention U-Net ensemble for interpretable polyp and instrument segmentation

The difficulty associated with screening and treating colorectal polyps alongside other gastrointestinal pathology presents an opportunity to incorporate computer-aided systems. This paper develops a deep learning pipeline that accurately segments colorectal polyps and various instruments used during endoscopic procedures. To improve transparency, we leverage the Attention U-Net architecture, enabling visualisation of the attention coefficients to identify salient regions. Moreover, we improve performance by incorporating transfer learning using a pre-trained encoder, together with test-time augmentation, softmax averaging, softmax thresholding and connected component labeling to further refine predictions.

Automatic Segmentation of Specific Intervertebral Discs through a Two-Stage MultiResUNet Model

The automatic segmentation of intervertebral discs from medical images is an important task for an intelligent clinical system. In this study, a deep learning model based on the MultiResUNet model for the automatic segmentation of specific intervertebral discs is presented. MultiResUNet can easily segment all intervertebral discs in MRI images; however, when only certain specific intervertebral discs need to be segmented, problems with segmentation errors, misalignment, and noise occur. In order to solve these problems, a two-stage MultiResUNet model is proposed. Connected-component labeling, automatic cropping, and distance transform are used in the proposed method. The experimental results show that the segmentation errors and misalignments of specific intervertebral discs are greatly reduced, and the segmentation accuracy is increased to about 94%. The performance of the proposed method proves its usefulness for the automatic segmentation of specific intervertebral discs over other deep learning models, such as the U-Net, CNN-based, Attention U-Net, and MultiResUNet models.

Comparative Study of Data Matrix Codes Localization and Recognition Methods

We provide a comprehensive and in-depth overview of the various approaches applicable to the recognition of Data Matrix codes in arbitrary images. All presented methods use the typical “L” shaped Finder Pattern to locate the Data Matrix code in the image. Well-known image processing techniques such as edge detection, adaptive thresholding, or connected component labeling are used to identify the Finder Pattern. The recognition rate of the compared methods was tested on a set of images with Data Matrix codes, which is published together with the article. The experimental results show that methods based on adaptive thresholding achieved a better recognition rate than methods based on edge detection.

Komputasi Budaya Untuk Pencarian Gambar Semantik Pada Lukisan Budaya Indonesia Dengan Deteksi Dan Informasi Aliran Lukisan

Lukisan merupakan salah satu gambaran kompleks yang mencerminkan pengamatan dan perasaan seniman terhadap lingkungan. Kondisi ini memperluas kebutuhan akan sistem pendeteksi citra budaya karena masyarakat awam yang kurang memiliki pengalaman artistik akan sulit mendapatkan kesan lukisannya. Oleh karena itu, peneliti menekankan penerapan lukisan budaya Indonesia ke dalam aplikasi mobile. Sistem yang diusulkan telah diimplementasikan pada 239 lukisan budaya Indonesia yang terdiri dari lima kategori gaya lukisan. Kategorinya adalah abstraksionisme, naturalisme, ekspresionisme, realisme, dan romantisme. Sistem mengekstrak 3 fitur, yaitu fitur warna, bentuk, dan tekstur. Ekstraksi ciri warna menggunakan Histogram 3D Color Vector Quantization. Ekstraksi fitur bentuk menggunakan Connected Component Labeling Algorithm (CCL) dengan menghitung nilai area, diameter setara, luas, convex hull, soliditas, eksentrisitas, dan perimeter masing-masing objek. Ekstraksi fitur tekstur menggunakan Gabor Transformation dengan 40 kernel. Sedangkan untuk ekstraksi impresi dilakukan survey terhadap beberapa orang tentang impresi lukisan budaya Indonesia. Survei ini dilakukan terhadap responden yang memahami seni lukis seperti pelukis, pemerhati lukisan, dan orang-orang yang berkecimpung di dunia seni rupa. Untuk menunjukkan gaya lukisan peneliti menggunakan proses klasifikasi menggunakan K-Nearest Neighbor. Hasil eksperimen menunjukan fitur warna sebagai fitur terbaik dalam impression query

Improved Connected Component Labeling Algorithm for Remote Sensing Image Classification

The main objective of this paper is to carry out a detailed analysis of the most popular Connected Component Labeling (CCL) algorithms for remote sensing image classification. This algorithm searches line-by-line, top to bottom to assign a splotch label to each current pixel that is connected to a splotch. This paper presents two new strategies that can be used to greatly improve the speed of connected component labeling algorithms. It assigns a label to a new object, most labeling algorithms use a scanning step that examines some of its neighbors. The first strategy deeds the dependencies among the neighbors to reduce the number of neighbors examined. The second strategy uses an array to store the equivalence information among the labels. This replaces the pointer based deep rooted trees used to store the same equivalence information. It reduces the memory required and also produces consecutive final labels. The connected component labeling assigns labels to a pixel such that adjacent pixels of the same features are assigned the same label. The paper presents a modification of this algorithm that allows the resolution of merged labels and experimental results demonstrate that proposed method is much more efficient than conventional methods for various kinds of color images. This method is improving the labeling algorithms and also benefits for other applications in computer vision and pattern recognition

Hijaiyah Letter Segmentation Using Connected Component Labeling Method

Arabic letters or Hijaiyah letters recognition is a challenge in itself because one letter consists of more than one character, namely the main character, companion character such as dots and lines, and punctuation called harakat. The image segmentation process is the most important in a character recognition system because it affects the separation of objects in an image. In this research, Hijaiyah letter segmentation aims to separate the letters according to the character of each letter using the Connected Component Labeling (CCL) method. Merging labels on each character will be done by looking for the Euclidean distance value from adjacent centroids. The experiment succeeded in segmenting each Hijaiyah character with an accuracy value of 86%.