Fast Segmentation Research Articles

EVALUATING MACHINE LEARNING MODELS FOR OPTIMAL CUSTOMER SEGMENTATION IN BANKING: A COMPARATIVE STUDY

This study presents a comparative analysis of machine learning algorithms for customer segmentation in the banking sector, utilizing a comprehensive dataset that includes transactional, demographic, and engagement attributes. Various clustering models, including K-Means, Gaussian Mixture Models (GMM), Hierarchical Clustering, DBSCAN, and Spectral Clustering, were evaluated to identify the most effective approach in terms of segmentation accuracy, scalability, and interpretability. The results revealed that Spectral Clustering consistently outperformed other models, offering superior accuracy and valuable insights into customer interactions across multiple banking touchpoints. While K-Means delivered fast and scalable segmentation, it lacked the flexibility needed for non-spherical clusters. The study also highlighted the benefits of a multi-dimensional dataset approach, which provided deeper insights into customer behavior, engagement, and loyalty. Although limitations such as computational complexity and scalability challenges remain, future research should focus on real-time data integration and multi-channel interactions across banking operations. This research not only contributes to machine learning applications in banking but also offers actionable strategies for targeted marketing, personalized customer engagement, risk management, and overall service optimization.

DETECTION OF DIABETIC RETINOPATHY FROM RETINAL FUNDUS IMAGES USING IMAGE SEGMENTAION AND AUTOMATIC DEEP LEARNING MODEL

Early diagnosis and treatment of diabetic retinopathy (DR) are critical to minimize or avoid vision loss. Furthermore, automatically locating the areas of the retinal image that may have lesions can help experts in the detection process. This paper proposes a deep learning technique with an adequate image segmentation mechanism for recognizing DR from retinal fundus data. Initially, the fundus data is collected from the openly available dataset. Then, the Fast K-Nearest Neighbour Image Segmentation Module (FKISM) is developed to segment the retinal lesions from the collected fundus data. Then, Auto Feature Selection Convolution Neural Network (AFSCNN) is applied to perform automated feature, earning and classification of DR. Simulations are done to analyze the proposed system’s effectiveness with the existing methodologies regarding classification metrics, and our proposed method illustrates superior performance by attaining higher outcomes in both training and testing contrasted to the traditional methods.

Fast segment anything model applied in a low-cost camera-based system for oyster growth monitoring

The increase in the global population has been accompanied by an increase in demand for animal protein, consequently driving the growth of the aquaculture sector as a sustainable solution to satisfy this demand. Observing the growth of animals in farm environments is vital to assess the efficiency of feeding methods, identify health problems or environmental stressors, and help implement sustainable practices. The application of cameras enables remote monitoring of various environments and activities in real time, providing quick insights, improving security measures, and facilitating decision-making based on the collected data. This paper proposes a low-cost camera system and an image processing-based strategy to monitor and measure the size of Pacific oysters inside a tank. The fast segment anything model algorithm is implemented in the Python programming language to perform image segmentation and generate object masks for subsequent processing. The size of oysters is obtained by measuring the distance between the parallel lines of the bounding box that is generated surrounding them. The largest absolute errors and relative errors for the estimates were 3 mm and 5.13%, respectively, when compared to measurements made manually. Furthermore, the width and height have standard deviations of 0.6 and 0.3 mm, respectively, indicating low data dispersion and good repeatability of the algorithm. Since the accuracy of the human eye in estimating a dimension without the aid of a distance measuring instrument tends to be lower than that obtained by the proposed tool, it is believed that the approach can significantly contribute to the management of oyster farm production, providing strategic insights based on animal growth for producers.

DeepNeXt: a lightweight polyp segmentation algorithm based on multi-scale attention.

Fast and accurate automatic segmentation of polyps in colonoscopy plays a crucial role in the early diagnosis and treatment of colon cancer. However, the current polyp segmentation algorithms based on deep neural networks suffer from the problems of larger models and lower segmentation accuracy. Meanwhile, achieving accurate segmentation of polyps is to improve the diagnostic efficiency of doctors, and this need motivates us to develop a set of lightweight models so that it can be easily embedded in clinical devices to meet the requirements of practical applications. This study aims to provide effective technical support for the rapid and precise segmentation of polyps in clinical applications. DeepNeXt, an innovative polyp segmentation model grounded in multi-scale attention mechanisms. DeepNeXt incorporates a multi-stage, lightweight convolutional encoder module, leveraging several lightweight convolutional layers for efficient and accurate feature extraction. Furthermore, it features a novel multi-stage feature fusion structure designed to circumvent the potential loss of feature information during the encoding phase. Additionally, the model employs a multi-scale attentional feature encoding module that harnesses multi-branch deep strip convolution techniques to extract multi-dimensional information from the feature maps post-encoding, thereby enhancing the neural network's capability to extract diverse feature information. Experimental validation on the Kvasir Segmentation Dataset (Kvasir-SEG dataset) and Colorectal Cancer-Clinic Datasetbase (CVC-ClinicDB datasets) demonstrates that DeepNeXt outperforms mainstream networks such as U-net, U-net++, TransUnet, SwinUnet, and TGANet in terms of parameters and floating-point operations (FLOPs). DeepNeXt achieved a FLOPs metric of only 3.04 G and a parameters (Params) metric of just 1.51 M, while delivering exceptional performance in segmentation. On the Kvasir-SEG dataset, it reached a mean intersection over union (mIOU) of 83.91, and on the CVC-ClinicDB dataset, 87.37. Additionally, the Dice and Recall metrics also showed superior results, highlighting that DeepNeXt strikes an optimal balance between computational efficiency, model compactness, and segmentation accuracy. In conclusion, we have proposed the DeepNeXt network, a novel lightweight multi-scale attention segmentation network tailored for computationally limited medical devices, which provides strong support for accurate and efficient polyp segmentation in clinical applications.

Lightweight video object segmentation: Integrating online knowledge distillation for fast segmentation

The typical shortcoming of STM (Space-Time Memory Network) mode video object segmentation algorithms is their high segmentation performance coupled with slow processing speeds, which poses challenges in meeting real-world application demands. In this work, we propose using an online knowledge distillation method to develop a lightweight video segmentation algorithm based on the STM mode, achieving fast segmentation while maintaining performance. Specifically, we utilize a novel adaptive learning rate to tackle the issue of inverse learning during distillation. Subsequently, we introduce a Smooth Block mechanism to reduce the impact of structural disparities between the teacher and student models on distillation outcomes. Moreover, to reduce the fitting difficulty of the student model on single-frame features, we design the Space-Time Feature Fusion (STFF) module to provide appearance and position priors for the feature fitting process of each frame. Finally, we employ a simple Discriminator module for adversarial training with the student model, to encourage the student model to learn the feature distribution of the teacher model. Extensive experiments show that our algorithm attains performance comparable to the current state-of-the-art on both DAVIS and YouTube datasets, despite running up to ×4 faster, with ×20 fewer parameters and ×30 fewer GFLOPS.

The faster the better? Comparing sustainability performance information in fast and luxury fashion

PurposeFashion brands, including fast and luxury segments, receive harsh criticism for engaging in unethical practices such as poor working conditions and environmental damage. As a result, fashion supply chains are pressured by stakeholders to publicly disclose internal supply chain performance information and to show a high level of supply chain transparency. This paper compares supply chain transparency in fast and luxury fashion in Europe.Design/methodology/approachBy applying the maturity curve of fashion supply chain transparency, developed by Muratore and Marques (2022), the websites of 20 fast and 20 luxury fashion brands were analysed and classified as Opaque, Translucent or Transparent.FindingsDespite its reputation, fast fashion demonstrated higher levels of transparency than luxury fashion. Luxury fashion only performed better in terms of the accessibility of sustainability information. Luxury brands avoided disclosing key transparency information, suggesting that they may be operating in contradiction to that which is inferred on their websites.Originality/valueThe findings of the study shed light on the sustainability credentials of the fashion industry, which has the potential to influence the purchase intentions of consumers, particularly millennials and Generation Z. Implications for practice are developed to highlight how fashion can improve its supply chain transparency.

A level-set method for fast image segmentation based on local pre-fitting and bilateral filtering

PurposeAlthough many conventional level-set approaches can be used for segmenting images containing factors such as noise and intensity inhomogeneities, they still can impact the accuracy of the results seriously. To solve this problem, a level-set method for fast image segmentation based on pre-fitting and bilateral filtering is proposed.Design/methodology/approachFirstly, an improved bilateral filter was investigated for image preprocessing. Secondly, by computing the local average intensity of the preprocessed enhanced picture, two local pre-fitting functions were defined. Thirdly, a new level-set energy functional was defined. Finally, a new distance regularized energy term based on the logarithmic and polynomial functions is proposed to evolve the level-set function in a smooth state.FindingsThe experimental results demonstrate that the proposed model has an excellent segmentation capability for images with noise and intensity inhomogeneities and has different degrees of performance improvement compared with the mainstream models.Originality/value(C1) An improved bilateral filter was investigated and integrated into the model. (C2) Proposing two local pre-fitting functions by computing the local average intensity of the preprocessed enhanced image. (C3) Proposing a new level-set energy functional. (C4) A new distance regularized energy term based on the logarithmic and polynomial functions is proposed to evolve the level set function in a smooth state. (C5) Analyzing and comparing the performance of the proposed model with other similar models.

A deep learning method for pointer meter reading recognition in inspection robots at refrigeration stations

Abstract To improve the intelligent operation and maintenance of large public building refrigeration stations, various automatic reading methods for pointer meters used in inspection robots have been developed. However, most methods exhibit low detection accuracy for pointer meters in refrigeration stations and poor segmentation of pointers and scale lines, ultimately leading to inaccurate readings. To address the challenges posed by the complex environments of refrigeration stations, this study proposes a deep learning-based image recognition method for precise pointer meter readings. Firstly, the pointer meters in the environment are detected and located by the YOLOV8-CBAM-Wise-IoU model. Secondly, an improved Fast Semantic Segmentation Network (Fast-SCNN) is used for image segmentation of the pointers and scale lines of the pointer meters, and a perspective transformation method is employed to correct images of tilted meters. Then, the image is transformed into a rectangular image using polar coordinate transformation. Finally, the accurate reading is calculated using the distance method, based on the relative distances between the pointers and scale lines. This research conducted experimental tests using an inspection robot in the actual environment of a large public building’s refrigeration station. The results demonstrate that the proposed method for pointer meter reading recognition achieves a fiducial error within 0.41%, which can realize the inspection task of the pointer meter in refrigeration stations.

TopoPIS: Topology-constrained pipe instance segmentation via adaptive curvature convolution

Precise and fast pipe instance segmentation is a critical component in industrial automatic assembly, facilitating accurate object detection and pose estimation, optimizing and supervising the assembly process. However, this problem is challenging due to topological errors on fine-scale structures caused by the pipes being complex and slender. To address these challenges, we propose a topology-constrained pipe instance segmentation network (TopoPIS) for complex stacking scene to achieve accurate segmentation with topological correctness. To better extract the features of complex and variable morphological pipes, adaptive curvature convolution is introduced to dynamically adapt to the slender pipe structure and capture critical features. To handle topological errors like broken connections, we propose a novel topological constraint loss function based on persistent homology, which greatly improves the topological continuity of the segmentation. Experimental results on real-world and unseen datasets demonstrate that our TopoPIS outperforms other methods regrading segmentation accuracy and topological continuity.

Dual convolution-transformer UNet (DCT-UNet) for organs at risk and clinical target volume segmentation in MRI for cervical cancer brachytherapy

Objective. MRI is the standard imaging modality for high-dose-rate brachytherapy of cervical cancer. Precise contouring of organs at risk (OARs) and high-risk clinical target volume (HR-CTV) from MRI is a crucial step for radiotherapy planning and treatment. However, conventional manual contouring has limitations in terms of accuracy as well as procedural time. To overcome these, we propose a deep learning approach to automatically segment OARs (bladder, rectum, and sigmoid colon) and HR-CTV from female pelvic MRI.Approach. In the proposed pipeline, a coarse multi-organ segmentation model first segments all structures, from which a region of interest is computed for each structure. Then, each organ is segmented using an organ-specific fine segmentation model separately trained for each organ. To account for variable sizes of HR-CTV, a size-adaptive multi-model approach was employed. For coarse and fine segmentations, we designed a dual convolution-transformer UNet (DCT-UNet) which uses dual-path encoder consisting of convolution and transformer blocks. To evaluate our model, OAR segmentations were compared to the clinical contours drawn by the attending radiation oncologist. For HR-CTV, four sets of contours (clinical + three additional sets) were obtained to produce a consensus ground truth as well as for inter/intra-observer variability analysis.Main results. DCT-UNet achieved dice similarity coefficient (mean ± SD) of 0.932 ± 0.032 (bladder), 0.786 ± 0.090 (rectum), 0.663 ± 0.180 (sigmoid colon), and 0.741 ± 0.076 (HR-CTV), outperforming other state-of-the-art models. Notably, the size-adaptive multi-model significantly improved HR-CTV segmentation compared to a single-model. Furthermore, significant inter/intra-observer variability was observed, and our model showed comparable performance to all observers. Computation time for the entire pipeline per subject was 12.59 ± 0.79 s, which is significantly shorter than the typical manual contouring time of >15 min.Significance. These experimental results demonstrate that our model has great utility in cervical cancer brachytherapy by enabling fast and accurate automatic segmentation, and has potential in improving consistency in contouring. DCT-UNet source code is available athttps://github.com/JHU-MICA/DCT-UNet.

LS-Net: lightweight segmentation network for dermatological epidermal segmentation in optical coherence tomography imaging.

Optical coherence tomography (OCT) can be an important tool for non-invasive dermatological evaluation, providing useful data on epidermal integrity for diagnosing skin diseases. Despite its benefits, OCT's utility is limited by the challenges of accurate, fast epidermal segmentation due to the skin morphological diversity. To address this, we introduce a lightweight segmentation network (LS-Net), a novel deep learning model that combines the robust local feature extraction abilities of Convolution Neural Network and the long-term information processing capabilities of Vision Transformer. LS-Net has a depth-wise convolutional transformer for enhanced spatial contextualization and a squeeze-and-excitation block for feature recalibration, ensuring precise segmentation while maintaining computational efficiency. Our network outperforms existing methods, demonstrating high segmentation accuracy (mean Dice: 0.9624 and mean IoU: 0.9468) with significantly reduced computational demands (floating point operations: 1.131 G). We further validate LS-Net on our acquired dataset, showing its effectiveness in various skin sites (e.g., face, palm) under realistic clinical conditions. This model promises to enhance the diagnostic capabilities of OCT, making it a valuable tool for dermatological practice.

PA-SLAM: Fast and Robust Ground Segmentation and Optimized Loop Detection Laser SLAM System

PA-SLAM: Fast and Robust Ground Segmentation and Optimized Loop Detection Laser SLAM System

A novel locally statistical active contour model for SAR image segmentation and faster algorithm

Abstract In this paper, a novel locally statistical active contour model (LACM) based on the Aubert-Aujol (AA) denoising model and variational level set method is proposed, which can be used for SAR image segmentation with intensity inhomogeneity. By simple transformation, a fast segmentation algorithm with a global optimization solver can be obtained. Experiments using some challenging synthetic images and Envisat SAR images demonstrate the superiority of our proposed models with respect to classic models. Further, we can find that the structure of the proposed algorithm is the same as that of the Transformer or RNN. We will study the nature of the above two types of algorithms to propose a more effective segmentation algorithm based on the energy model.

A Comparison of Throwing Arm Kinetics and Ball Velocity in High School Pitchers With Overall Fast and Overall Slow Cumulative Joint and Segment Velocities

Background: Individual maximum joint and segment angular velocities have shown positive associations with throwing arm kinetics and ball velocity in baseball pitchers. Purpose: To observe how cumulative maximum joint and segment angular velocities, irrespective of sequence, affect ball velocity and throwing arm kinetics in high school pitchers. Study Design: Descriptive laboratory study. Methods: High school (n = 55) pitchers threw 8 to 12 fastball pitches while being evaluated with 3-dimensional motion capture (480 Hz). Maximum joint and segment angular velocities (lead knee extension, pelvis rotation, trunk rotation, shoulder internal rotation, and forearm pronation) were calculated for each pitcher. Pitchers were classified as overall fast, overall slow, or high velocity for each joint or segment velocity subcategory, or as population, with any pitcher eligible to be included in multiple subcategories. Kinematic and kinetic parameters were compared among the various subgroups using t tests with post hoc regressions and multivariable regression models created to predict throwing arm kinetics and ball velocity, respectively. Results: The lead knee extension and pelvis rotation velocity subgroups achieved significantly higher normalized elbow varus torque (P = .016) and elbow flexion torque (P = .018) compared with population, with equivalent ball velocity (P = .118). For every 1-SD increase in maximum pelvis rotation velocity (87 deg/s), the normalized elbow distractive force increased by 4.7% body weight (BW) (B = 0.054; β = 0.290; P = .013). The overall fast group was older (mean ± standard deviation, 16.9 ± 1.4 vs 15.4 ± 0.9 years; P = .007), had 8.9-mph faster ball velocity (32.7 ± 3.1 vs 28.7 ± 2.3 m/s; P = .002), and had significantly higher shoulder internal rotation torque (63.1 ± 17.4 vs 43.6 ± 12.0 Nm; P = .005), elbow varus torque (61.8 ± 16.4 vs 41.6 ± 11.4 Nm; P = .002), and elbow flexion torque (46.4 ± 12.0 vs 29.5 ± 6.8 Nm; P < .001) compared with the overall slow group. A multiregression model for ball velocity based on maximum joint and segment angular velocities and anthropometrics predicted 53.0% of variance. Conclusion: High school pitchers with higher maximum joint and segment velocities, irrespective of sequence, demonstrated older age and faster ball velocity at the cost of increased throwing shoulder and elbow kinetics. Clinical Relevance: Pitchers and coaching staff should consider this trade-off between faster ball velocity and increasing throwing arm kinetics, an established risk factor for elbow injury.

Machine learning driven image segmentation and shape clustering of algal microscopic images obtained from various water types

Algae and cyanobacteria are microorganisms found in almost all fresh and marine waters, where they can pose environmental and public health risks when they grow excessively and produce blooms. Accurate identification and quantification of these microorganisms are vital for ecological research, water quality monitoring, and public health safety. However, traditional methods of manually counting and morphologically identifying these microorganisms are time-consuming and prone to human error. Application of the machine learning-driven Fast Segment Anything Model (FastSAM), an image segmentation model, automates and potentially enhances the accuracy and efficiency of cell identification and enumeration from microscopic images. We assessed FastSAM for algal cell image segmentation, and three clustering evaluation metrics. Segmentation of microscopic images of algal and cyanobacterial cells in water and treated wastewater samples using the FastSAM algorithm, which is a Vision Transformer (ViT) based algorithm that utilizes a compact ViT encoder, demonstrated benefits and challenges of this machine learning driven image processing. Notably, the pre-trained algorithm segmented entire elements in all microscopic images used in this study. Depending on the shape, 50–100 % similarity was observed between machine-based segmentation and manual validation of all segmented elements, with 100 % of single cells being correctly segmented by FastSAM. The performance of clustering metrics varied between 57–94 % with the Spectral Angle Mapper achieving the most accurate performance, 84–94 %, compared to the manually chosen clustering benchmarks. Cyanobacterial and algal communities are biologically diverse and have ecological significance. The application of image clustering techniques in studying their cell shapes marks an important advancement in microbial ecology and environmental monitoring. As technology progresses, these methods will become increasingly utilised to decipher the complex roles that algae and cyanobacteria play in our ecosystems supporting mitigation and public health protection measures.

Image Multi-threshold Segmentation Based on an Ameliorated Harmony Search Optimization Algorithm

Image segmentation is the basis and premise of image processing, though traditional multi-threshold image segmentation methods are simple and effective, they suffer the problems of low accuracy and slow convergence rate. For that reason, this paper introduces the multi-threshold image segmentation scheme by combining the harmony search (HS) optimization algorithm and the maximum between-class variance (Otsu) to solve them. Firstly, to further improve the performance of the basic HS, an ameliorated harmony search (AHS) is put forward by modifying the generation method of the new harmony improvisation and introducing a convergence coefficient. Secondly, the AHS algorithm, which takes the maximum between-class variance as its objective function, namely AHS-Otsu, is applied to image multi-level threshold segmentation. Finally, six test images are selected to verify the multilevel segmentation performance of AHS-Otsu. Peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) are two commonly used metrics for evaluating the effectiveness of image segmentation, which are both used in this article. Comprehensive experimental results indicate that the AHS-Otsu does not only has fast segmentation processing speed, but also can obtain more accurate segmentation performance than others, which prove the effectiveness and potential of the AHS-Otsu algorithm in the field of image segmentation especially for the multi-threshold.

Fast supervoxel segmentation of connectivity median simulation based on Manhattan distance

Supervoxels provide a more natural and compact 3D point cloud representation, which is a fundamental problem in point cloud processing and has attracted extensive attention from many scholars. At present, although the supervoxel segmentation has achieved significant results, the processing efficiency still restricts its further processing in the face of large-scale point cloud processing. Therefore, this paper proposes a new supervoxel segmentation method for point clouds to achieve fast supervoxel over-segmentation of large-scale point clouds while maintaining good boundaries and accuracy. Firstly, the supervoxel segmentation is formulated as a subset selection problem. With the idea of greedy strategy, the interval quick sorting and region growing methods is designed to realize the rapid merging of subsets in the iterative process. Secondly, the Manhattan distance metric and the seed point median simulation method are proposed to enhance the precision of iterative optimization and improve the efficiency of point cloud supervoxel segmentation. Experimental results show that, compared with state-of-the-art supervoxel segmentation method, the proposed method achieves the best results under UE, GCE, BR and other indicators, while reducing the running time of the algorithm by 10.6% to 28.3%.

Fast and robust object region segmentation with self-organized lattice Boltzmann based active contour method

Fast and robust object region segmentation with self-organized lattice Boltzmann based active contour method

Segmentation and characterization of macerated fibers and vessels using deep learning

PurposeWood comprises different cell types, such as fibers, tracheids and vessels, defining its properties. Studying cells’ shape, size, and arrangement in microscopy images is crucial for understanding wood characteristics. Typically, this involves macerating (soaking) samples in a solution to separate cells, then spreading them on slides for imaging with a microscope that covers a wide area, capturing thousands of cells. However, these cells often cluster and overlap in images, making the segmentation difficult and time-consuming using standard image-processing methods.ResultsIn this work, we developed an automatic deep learning segmentation approach that utilizes the one-stage YOLOv8 model for fast and accurate segmentation and characterization of macerated fiber and vessel form aspen trees in microscopy images. The model can analyze 32,640 x 25,920 pixels images and demonstrate effective cell detection and segmentation, achieving a mAP0.5-0.95\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\hbox {mAP}_{0.5-0.95}$$\\end{document} of 78 %. To assess the model’s robustness, we examined fibers from a genetically modified tree line known for longer fibers. The outcomes were comparable to previous manual measurements. Additionally, we created a user-friendly web application for image analysis and provided the code for use on Google Colab.ConclusionBy leveraging YOLOv8’s advances, this work provides a deep learning solution to enable efficient quantification and analysis of wood cells suitable for practical applications.

Accurate neuron segmentation method for one-photon calcium imaging videos combining convolutional neural networks and clustering

One-photon fluorescent calcium imaging helps understand brain functions by recording large-scale neural activities in freely moving animals. Automatic, fast, and accurate active neuron segmentation algorithms are essential to extract and interpret information from these videos. One-photon imaging videos’ low resolution, high noise, and high background fluctuation pose significant challenges. Here, we develop a software pipeline to address the challenges of processing one-photon calcium imaging videos. We extend our previous two-photon active neuron segmentation algorithm, Shallow U-Net Neuron Segmentation (SUNS), to better suppress background fluctuations in one-photon videos. We also develop additional neuron extraction (ANE) to locate small or dim neurons missed by SUNS. To train our segmentation method, we create ground truth neurons by developing a manual labeling pipeline assisted with semi-automatic refinement. Our method is more accurate and faster than state-of-the-art techniques when processing simulated videos and multiple experimental datasets acquired over various brain regions with different imaging conditions.