Frame Segmentation Research Articles

Biaxial Resistance of Pre-Engineered Beam Hangers in Glulam

In timber construction, Glulam post-and-beam systems are commonly used to transfer vertical loads to the foundation. In such systems, the connections play a critical role in structural performance. Pre-engineered connectors, which facilitate fast and efficient assembly, are typically designed to resist only vertical shear loads. However, during seismic and wind events, post-and-beam systems deform horizontally, and axial forces develop at the connections. In this research, the performance of RICON and MEGANT pre-engineered connectors was studied under biaxial loading involving concurrent shear and axial forces. A total of 12 full-scale tests on Glulam frame segments were conducted. Neither type of connector experienced any resistance loss under concurrent shear loads equal to the factored shear resistance and axial loads equal to 5% of the factored shear resistance. The axial load-carrying capacity of the RICON and MEGANT connectors was up to 124% and 97% of their factored shear resistance, respectively. The global failure of all the studied connectors demonstrated both ductility and residual deformation capacity. These results provide valuable information for engineers designing Glulam post-and-beam systems in seismic regions.

Effectiveness of Comprehensive Video Datasets: Toward the Development of an Artificial Intelligence Model for Ultrasonography-Based Severity Diagnosis of Carpal Tunnel Syndrome.

Advances in diagnosing carpal tunnel syndrome (CTS) using ultrasonography (US) and artificial intelligence (AI) aim to replace nerve conduction studies. However, a method for accurate severity diagnosis remains unachieved. We explored the potential of comprehensive video data formats for constructing an effective model for diagnosing CTS severity. We studied 75 individuals (52 with CTS) from 2019 to 2022, categorizing them into 3 groups based on disease severity. We recorded 132 US videos of carpal tunnel during finger movement. Features of the median nerve (MN) were extracted from automatically segmented US video frames, from which 3 datasets were created: a comprehensive video dataset with full information, a key metrics dataset, and an initial frame dataset with the least information. We compared the accuracy of machine learning algorithms for classifying CTS severity into 3 groups across these datasets using 63-fold cross-validation. The cross-sectional area of the MN correlated with severity (P < .05) but MN displacement did not. The algorithm using the comprehensive video dataset exhibited the highest sensitivity (1.00) and accuracy (0.75). Our study demonstrated that utilizing comprehensive video data enables a more accurate US-based diagnosis of CTS severity. This underscores the value of capturing the patterns of MN deformation and movement, which cannot be captured by representative metrics such as medians or maximums. By further developing an AI model based on our findings, a simpler and painless method for assessing CTS severity can be achieved.

A reproducible representation of healthy tibiofemoral kinematics during stair descent using REFRAME – Part II: Exploring optimisation criteria and inter-subject differences

Kinematic analysis is a central component of movement biomechanics, describing the relative motion of joint segments during different activities, in different subject cohorts, and at different timepoints. Establishing whether two sets of kinematic signals represent fundamentally similar or different underlying motion patterns is especially challenging, given 1) the lack of consensus around reference frame and joint axis definition, and 2) the substantial effect that minimal variations in frame position and orientation are known to have on signal magnitude and characteristics. As such, enormous variability in the reporting of tibiofemoral kinematics has resulted in joint movement patterns that remain controversially discussed. Previously, we demonstrated the ability of the REference FRame Alignment MEthod (REFRAME) to reorientate and reposition differently aligned local segment frames to achieve convergence in signals representing the same underlying motion, thereby offering a novel approach to consistently report joint motion. In this study, for the first time, we apply REFRAME to assess the rotational and translational in vivo tibiofemoral motion of ten healthy subjects during stair descent based on kinematic signals collected using a moving videofluoroscope. Kinematics were analysed before and after different REFRAME implementations, revealing generally neutral ab/adduction behaviour, accompanied by varying degrees of a sinusoidal int/external tibial rotation pattern over the activity cycle. Our data demonstrate that different selected implementations of REFRAME are able to highlight different characteristics of the motion patterns: Minimisation of the translational root-mean-square revealed proximodistal translation patterns with overall neutral progression, while anteroposterior translation showed seemingly different levels of correlation with flexion/extension in different subjects. On the other hand, REFRAME minimisation of translational variances exposed differences in the relative mean displacement between the femoral and tibial origins between subjects, highlighting differences in mean centre of rotation positions. This early application of REFRAME for providing an understanding of tibiofemoral kinematics demonstrates the potential of this novel approach to bring clarity to an otherwise complex representation of highly variable time-series signals, while highlighting the philosophical challenges of clinically interpretating kinematic signals in the first place.

Exploiting Temporal Features in Calculating Automated Morphological Properties of Spiky Nanoparticles Using Deep Learning

Object segmentation in images is typically spatial and focuses on the spatial coherence of pixels. Nanoparticles in electron microscopy images are also segmented frame by frame, with subsequent morphological analysis. However, morphological analysis is inherently sequential, and a temporal regularity is evident in the process. In this study, we extend the spatially focused morphological analysis by incorporating a fusion of hard and soft inductive bias from sequential machine learning techniques to account for temporal relationships. Previously, spiky Au nanoparticles (Au-SNPs) in electron microscopy images were analyzed, and their morphological properties were automatically generated using a hourglass convolutional neural network architecture. In this study, recurrent layers are integrated to capture the natural, sequential growth of the particles. The network is trained with a spike-focused loss function. Continuous segmentation of the images explores the regressive relationships among natural growth features, generating morphological statistics of the nanoparticles. This study comprehensively evaluates the proposed approach by comparing the results of segmentation and morphological properties analysis, demonstrating its superiority over earlier methods.

Driveway Detection for Weed Management in Cassava Plantation Fields in Thailand Using Ground Imagery Datasets and Deep Learning Models

Weeds reduce cassava root yields and infest furrow areas quickly. The use of mechanical weeders has been introduced in Thailand; however, manually aligning the weeders with each planting row and at headland turns is still challenging. It is critical to clear weeds on furrow slopes and driveways via mechanical weeders. Automation can support this difficult work for weed management via driveway detection. In this context, deep learning algorithms have the potential to train models to detect driveways through furrow image segmentation. Therefore, the purpose of this research was to develop an image segmentation model for automated weed control operations in cassava plantation fields. To achieve this, image datasets were obtained from various fields to aid weed detection models in automated weed management. Three models—Mask R-CNN, YOLACT, and YOLOv8n-seg—were used to construct the image segmentation model, and they were evaluated according to their precision, recall, and FPS. The results show that YOLOv8n-seg achieved the highest accuracy and FPS (114.94 FPS); however, it experienced issues with frame segmentation during video testing. In contrast, YOLACT had no segmentation issues in the video tests (23.45 FPS), indicating its potential for driveway segmentation in cassava plantations. In summary, image segmentation for detecting driveways can improve weed management in cassava fields, and the further automation of low-cost mechanical weeders in tropical climates can be performed based on the YOLACT algorithm.

Lightweight hashing with contrastive self-cross XLNet and combinational similarity matching for content-based video retrieval

ABSTRACT Text-to-video retrieval systems enable simple and natural video searches, but existing methods often employ single queries with noisy annotation and low-quality descriptions. A novel lightweight hashing with Contrastive self-cross XLNet and Combinational similarity matching is proposed to provide high-quality descriptions and clarity in annotation. Issues in existing text and video processing approaches, such as redundancy in keyword generation and extracting frames without considering optical flow estimation of keyframes, are solved using a novel Textual Labelling and Automated key frame segmentation. The 3D Cuboid Net with Lightweight Hashing is proposed, which selects features and converts them into hash codes using Lightweight LSS-Net based hashing, reducing memory utilization. Additionally, a novel Combinational Attentive Text-Video Retrieval is proposed, which eliminates multiple narrow extrema matching error surfaces and increases retrieval validity. The model efficiently solves video retrieval problems with high accuracy, recall, and sensitivity. This novel approach addresses the limitations of existing text-to-video retrieval methods and improves the overall quality of the retrieval process.

Real-time path planning for autonomous vehicle off-road driving.

Autonomous driving is a growing research area that brings benefits in science, economy, and society. Although there are several studies in this area, currently there is no a fully autonomous vehicle, particularly, for off-road navigation. Autonomous vehicle (AV) navigation is a complex process based on application of multiple technologies and algorithms for data acquisition, management and understanding. Particularly, a self-driving assistance system supports key functionalities such as sensing and terrain perception, real time vehicle mapping and localization, path prediction and actuation, communication and safety measures, among others. In this work, an original approach for vehicle autonomous driving in off-road environments that combines semantic segmentation of video frames and subsequent real-time route planning is proposed. To check the relevance of the proposal, a modular framework for assistive driving in off-road scenarios oriented to resource-constrained devices has been designed. In the scene perception module, a deep neural network is used to segment Red-Green-Blue (RGB) images obtained from camera. The second traversability module fuses Light Detection And Ranging (LiDAR) point clouds with the results of segmentation to create a binary occupancy grid map to provide scene understanding during autonomous navigation. Finally, the last module, based on the Rapidly-exploring Random Tree (RRT) algorithm, predicts a path. The Freiburg Forest Dataset (FFD) and RELLIS-3D dataset were used to assess the performance of the proposed approach. The theoretical contributions of this article consist of the original approach for image semantic segmentation fitted to off-road driving scenarios, as well as adapting the shortest route searching A* and RRT algorithms to AV path planning. The reported results are very promising and show several advantages compared to previously reported solutions. The segmentation precision achieves 85.9% for FFD and 79.5% for RELLIS-3D including the most frequent semantic classes. While compared to other approaches, the proposed approach is faster regarding computational time for path planning.

Video object segmentation based on dynamic perception update and feature fusion

The current popular video object segmentation algorithms based on memory network indiscriminately update the frame information to the memory pool, fails to make reasonable use of the historical frame information, causing frame information redundancy in the memory pool, resulting in the increase of the computation amount. At the same time, the mask refinement method is relatively rough, resulting in blurred edges of the generated mask. To solve these problems, This paper proposes a video object segmentation algorithm based on dynamic perception update and feature fusion. In order to reasonably utilize the historical frame information, a dynamic perception update module is proposed to selectively update the segmentation frame mask. Meanwhile, a mask refinement module is established to enhance the detail information of the shallow features of the backbone network. This module uses a double kernels fusion block to fuse the different scale information of the features, and finally uses the Laplacian operator to sharpen the edges of the mask. The experimental results show that on the public datasets DAVIS2016, DAVIS2017 and YouTube-VOS18, the comprehensive performance of the algorithm in this paper reaches 86.9%, 79.3% and 71.6%, respectively, and the segmentation speed reaches 15FPS on the DAVIS2016 dataset. Compared with many mainstream algorithms in recent years, it has obvious advantages in performance.

Hierarchical segmentation of surgical scenes in laparoscopy.

Segmentation of surgical scenes may provide valuable information for real-time guidance and post-operative analysis. However, in some surgical video frames there is unavoidable ambiguity, leading to incorrect predictions of class or missed detections. In this work, we propose a novel method that alleviates this problem by introducing a hierarchy and associated hierarchical inference scheme that allows broad anatomical structures to be predicted when fine-grained structures cannot be reliably distinguished. First, we formulate a multi-label segmentation loss informed by a hierarchy of anatomical classes and then train a network using this. Subsequently, we use a novel leaf-to-root inference scheme ("Hiera-Mix") to determine the trade-off between label confidence and granularity. This method can be applied to any segmentation model. We evaluate our method using a large laparoscopic cholecystectomy dataset with 65,000 labelled frames. We observed an increase in per-structure detection F1 score for the critical structures, when evaluated across their sub-hierarchies, compared to the baseline method: 6.0% for the cystic artery and 2.9% for the cystic duct, driven primarily by increases in precision of 11.3% and 4.7%, respectively. This corresponded to visibly improved segmentation outputs, with better characterisation of the undissected area containing the critical structures and fewer inter-class confusions. For other anatomical classes, which did not stand to benefit from the hierarchy, performance was unimpaired. Our proposed hierarchical approach improves surgical scene segmentation in frames with ambiguity, by more suitably reflecting the model's parsing of the scene. This may be beneficial in applications of surgical scene segmentation, including recent advancements towards computer-assisted intra-operative guidance.

Robust Clustering-Based Automated Video Shot Boundary Detection Using Handcrafted and Deep Feature Fusion

Video shot boundary detection (VSBD) plays a key role in analyzing, summarizing, indexing and retrieving content-based data from videos. Many artificial intelligence (AI)-based techniques have recently been introduced to detect the gradual transition from video frames. However, those techniques fail to detect the presence of inter-classes like fade-in, fade-out and dissolve from gradual transition video frames. In addition, the existing techniques face high computational complexity during detecting transitions from the video frames. This research brings novel clustering-based techniques to classify the inter-classes like fade-in, fade-out and dissolve from gradual transition video frames. At the initial stage, color histogram differences (CHD) technique is introduced to detect the abrupt transition from the video frames. The identified abrupt transitions are completely removed from the video frames. Then, segmentation is done to segment the gradual transitions from the video frames. The segmented gradual transition frames are then given to extract the handcrafted and deep features from the video frames. The deep features are extracted from the segments using the Morlet Wavelet-assisted modified stacked autoencoder (MW-MSAE) technique. The extracted handcrafted features and deep features are then concatenated together, and finally, fused feature vectors are obtained. The fused features are then fed to the robust deep [Formula: see text]-means map clustering method (RDKMM) to aggregate based on similar features. For calculating the similar features, a similarity-based correlation calculation (SBCC) is done in adjacent frames to determine gradual shot transitions like fade-in, fade-out and dissolve. The dataset used in this research is the TREC Video Retrieval Evaluation (TRECVID) 2021 dataset. In the experimental scenario, an accuracy of 95.5%, a sensitivity of 95.3%, a specificity of 96.9%, a precision of 93.9%, an F-measure of 94.6% and Mathew’s correlation coefficient (MCC) of 92.4% are obtained.

Block based motion estimation model using CNN with representative point matching algorithm for object tracking in videos

Motion estimation is considered significant for tracking the movement of an object in video sequences, and it is widely used in various video processing applications. Traditionally, many researchers focus on pixel-based motion estimation for object tracking, but it experienced increased computation time and cost. To reduce computation time, the utilization of a block-based motion estimation approach for object tracking is a recent trend. The existing block-based approach faces difficulty in finding representative points within the intensity domain. Therefore, this current research merged the deep learning approach with a block-matching algorithm for achieving efficient object tracking. In this proposed work, initially, video sequences are collected from a benchmark video dataset. Then, the acquired video sequences are segmented into frames. From the segmented frames, current and previous frames are considered for motion estimation. Frames are sent for the data augmentation process in which the process of flipping, cropping, and rotation is carried out. Then, the augmented frames are sent into Convolutional Neural Network (CNN) for feature extraction. Representative Point Matching (RPM) is used to estimate the motion vector based on the extracted features. After estimating the motion vector, the similarity between two consecutive frames is found using Structural Similarity Index (SSIM) technique. Finally, based on the similarity score, the movement of an object in the video is tracked effectively. Simulation analysis of the proposed block-based motion estimation model is done by evaluating some performance metrics. RMSE, PSNR, Execution Time, SSIM, and accuracy obtained for the proposed model are 27.5, 26.5 db, 31 sec, 0.91, and 94 %. This analysis suggested that the proposed CNN-RPM motion estimation model performs better in tracking the movement of the object.

Frame Segmentation in Ultrahigh-Speed Spectrum Analyzer Based on Time-Lens

Frame Segmentation in Ultrahigh-Speed Spectrum Analyzer Based on Time-Lens

Automatic segmentation of curtain wall frame using a context collaboration pyramid network

Accurate positioning of curtain wall frames is crucial for the automated installation of curtain wall modules. However, the current robot-based installation methods overly depend on visual guidance from operators, resulting in high costs and limiting construction efficiency. The development of deep learning has introduced an image segmentation approach that offers a new solution for the visual positioning of curtain wall frames. This paper proposes a context collaboration pyramid network to automatically segment curtain wall frames by incorporating context interaction and channel guided pyramid structure. The model adopts an “encoder-decoder” architecture with a feature interaction block strategically inserted between the encoder and decoder. Specifically, the encoder utilizes the pyramid pooling Transformer as a backbone to extract multi-level features from original RGB images. The decoder employs a channel guided pyramid convolution module to integrate multi-scale features and achieve finer prediction. Meanwhile, a context interaction fusion module between the features of adjacent levels was designed carefully to enhance the collaboration of the architecture. In addition, a benchmark dataset for the curtain wall frame segmentation task, consisting of 1547 images, was established. The dataset incorporates challenging scenarios, including strong lights, low contrast, and cluttered backgrounds. This method is evaluated on the collected dataset, and achieves an impressive accuracy of 97.30% and an F1-Score of 88.95%, outperforming other segmentation networks. Overall, the proposed method can extract target information accurately and efficiently and provide critical visual guidance for the robot, so as to promote the automatic installation level of the curtain wall module.

Modulated Memory Network for Video Object Segmentation

Existing video object segmentation (VOS) methods based on matching techniques commonly employ a reference set comprising historical segmented frames, referred to as ‘memory frames’, to facilitate the segmentation process. However, these methods suffer from the following limitations: (i) Inherent segmentation errors in memory frames can propagate and accumulate errors when utilized as templates for subsequent segmentation. (ii) The non-local matching technique employed in top-leading solutions often fails to incorporate positional information, potentially leading to incorrect matching. In this paper, we introduce the Modulated Memory Network (MMN) for VOS. Our MMN enhances matching-based VOS methods in the following ways: (i) Introducing an Importance Modulator, which adjusts memory frames using adaptive weight maps generated based on the segmentation confidence associated with each frame. (ii) Incorporating a Position Modulator that encodes spatial and temporal positional information for both memory frames and the current frame. The proposed modulator improves matching accuracy by embedding positional information. Meanwhile, the Importance Modulator mitigates error propagation and accumulation by incorporating confidence-based modulation. Through extensive experimentation, we demonstrate the effectiveness of our proposed MMN, which also achieves promising performance on VOS benchmarks.

Noise signature identification using mobile phones for indoor localization

Indoor localization is still nowadays a challenge with room to improve. Even though there are many different approaches that have evidenced as effective, most of them require specific hardware or infrastructure deployed along the building that can be discarded in many potential scenarios. Others that do not require such on-site infrastructure, like inertial navigation-based systems, entail certain accuracy problems due to the accumulation of errors. However, this error-accumulation can be mitigated using beacons that support the recalibration of the system. The more frequently beacons are detected, the smaller will be the accumulated error. In this work, we evaluate the use of the noise signature of the rooms of a building to pinpoint the current location of a low-cost Android device. Despite this strategy is not a complete indoor localization system (two rooms could share the same signature), it allows us to generate beacons automatically. The noise recorded by the device is preprocessed performing audio filtering, audio frame segmentation, and feature extraction. We evaluated binary (determining if the ambient sound recording belonged to a specific room) and multi-class (identifying which room an ambient noise recording belonged to by comparing it amongst the remaining 18 rooms from the original 19 rooms sampled) classification methods. Our results indicate that the two Stacking techniques and K-Nearest Neighbor (KNN) machine learning classifier are the most successful methods in binary classification with an average accuracy of 99.19%, 99,08%, and 99.04%. In multi-class classification the average accuracy for KNN is 90.77%, and 90.52% and 90.15% for both Voting techniques.

Effective flame detection and classification using deep feature of QuickDenseNet and Ensemble Score Voted SVM

Fire poses a significant threat to both lives and property, necessitating effective early detection measures. Despite challenges in identifying smoke and fire in their initial stages, we have devised a cost-efficient visual detection system. Early fire detection enhances its potential effectiveness. CCTV surveillance systems are now commonplace in developed countries, serving as tools for periodic monitoring of various locations. However, fluctuating ambient light conditions, camera angles, and seasonal variations can introduce data distortions, occlusions, and impact model accuracy. To address these issues, we’ve implemented a method combining deep learning networks and machine learning strategies for flame detection and direction classification. Our innovative QuickDenseNet extracts dense features from segmented flame video frames. We introduce the Ensemble Score Voted SVM (ESV-SVM), employing SVM as the primary learner and score voting as the auxiliary learner. Our approach is rigorously evaluated through simulations, measuring accuracy and various Key Performance Indices (KPIs), including Precision, F1-score, Recall, Correlation, Error, FPR, and Correlation Coefficients. Remarkably, our proposed method achieves an impressive precision rate of approximately 99.5%.

Robust Object Segmentation using 3D Mesh Models and Self-Supervised Learning

The proposed method presents an innovative solution that addresses the challenges of unlabeled data, robust performance in unknown environments, and foreground-background differentiation in object segmentation. The objective of this work is to improve object segmentation tasks with an unannotated pool of data, derive model structure to demystify and educate large AI models by using knowledge distillation method, implementing framework with self-supervised learning analyse impacting and driving factors. Proposed method handles unlabeled data effectively, improving accuracy and generalization in diverse scenarios. It demonstrates remarkable reliability in unknown environments, ensuring consistent and accurate object segmentation in complex visual contexts. By incorporating novel techniques, the approach overcomes the longstanding problem of achieving consistent and non-trivial solutions for object segmentation, offering a comprehensive and effective solution for computer vision and image analysis. The literature review reveals that the DINO model using a ViT-B/8 backbone achieves an impressive 80.1% accuracy in linear regression, while the DINO model with a ViT-B/16 backbone attains the highest k-NN accuracy at 78.3%. The approach integrates deep neural networks for object segmentation with Kalman filtering for non-linear state estimation. Specifically, depth information obtained from point clouds is directly incorporated into the Unscented Kalman filtering, allowing for efficient and accurate performance. A one-parameter procedure is employed to identify and eliminate irregularities in point clouds, significantly improving the performance of segmenting objects. By incorporating self-supervised learning with distillation, the method effectively predicts and addresses the challenge of refining trivial solutions, leading to improved performance in viewpoint and illumination scenarios. Decision-making is deferred until a global view of the entire frame can be established, enabling accurate and consistent object segmentation in subsequent frames. The proposed method, SSL-DM+UKF, outperforms existing approaches in various challenges, including occlusion, viewpoint changes, and illumination variations. It achieves statistical superiority over the baseline SSL-Distillation Model by up to 2.50% in different scenarios. Incorporating depth observations through the Unscented Kalman Filter and using self-supervised learning with knowledge distillation are key factors contributing to its success. SSL-DM+UKF offers a valuable and efficient solution for real-time object segmentation in dynamic environments.

Deep-learning based analysis of intracoronary optical coherence tomography images: detection and characterization of lipid plaques

Abstract Background Intracoronary optical coherence tomography (OCT) enables in vivo detection and characterization of atherosclerotic disease. However, manual assessment of OCT images is time-consuming and subject to intra- and interobserver variability. Therefore, automated and trustworthy methods for plaque assessment are warranted. Purpose To develop and validate an artificial intelligence based algorithm for detection and characterization of lipid plaques on OCT images. Methods From the prospective, observational PECTUS-obs study, we studied a representative subset of OCT pullbacks performed in fractional flow reserve-negative non-culprit lesions in patients presenting with myocardial infarction. Manual segmentation of cross-sectional frames (e.g. lumen, tunica intima including the fibrous cap, tunica media and lipid pools) was performed by trained experts. Pullbacks were randomly divided into a training and test set. For automated segmentation and analysis, a two-dimensional no-new U-shaped neural network (nnUNet) was constructed with 5-fold cross validation. To test the diagnostic performance of the nnUnet for detection of lipid plaques, sensitivity, specificity and Cohen’s kappa were calculated. As for plaque characterization, the lipid arc and minimal fibrous cap thickness were measured manually in the test set on each frame containing a lipid plaque. Values of lipid arc and minimal fibrous cap thickness obtained following automated assessment were compared to manual analysis using the intraclass correlation coefficient (ICC) for absolute agreement. Results In total, 1215 frames were used for training and 162 frames for testing of the nnUNet. Lipid plaques were present in 77 frames (47.5%) in the test set. Substantial agreement (κ=0.79) for detection of a lipid plaque was achieved with a sensitivity and specificity of 96.1% and 83.5%, respectively. As for plaque characterization, we found good reliability for lipid arc assessment (ICC 0.88, mean difference 6±32º) and moderate reliability for assessment of the minimal fibrous cap thickness (ICC 0.54, mean difference 23±140µm). Conclusion We developed a deep learning algorithm for automated analysis of intracoronary OCT images, that was able to accurately detect lipid plaques. Moreover, the algorithm showed promising results in terms of plaque characterization. With further refinements, the developed algorithm has the potential to reduce the time required for OCT interpretation, ultimately enabling more efficient, real-time use of this imaging modality in clinical practice.

Comparative Analysis of U-Net and DeepLab for Automatic Polyp Segmentation in Colonoscopic Frames Using CVC-ClinicDB Dataset

In the context of medical image segmentation, accurate polyp detection in colonoscopy videos is crucial for early colorectal disease diagnosis. This study compares U-Net and DeepLab deep learning models in automatically segmenting polyps using the CVC-ClinicDB dataset. The dataset comprises 612 images from 31 colonoscopy sequences with a resolution of 384×288.Our primary metric is Mean Intersection over Union (IoU), measuring the overlap between predicted and ground truth masks. We also evaluate Mean Dice Loss for comprehensive comparison. U-Net demonstrates superior performance, with a Mean IoU Score of 0.9897 and a low Mean Dice Loss of 0.0523, indicating consistent and accurate polyp segmentation. In contrast, DeepLab achieves a Mean IoU Score of 0.9676 and a slightly higher Mean Dice Loss of 0.0417, showing good results but being outperformed by U-Net.In conclusion, U-Net excels in automatic polyp segmentation, offering high accuracy and robustness. These findings advance computer-aided diagnosis for colorectal diseases, potentially enhancing early and precise polyp detection.

Comparative Analysis of U-Net and DeepLab for Automatic Polyp Segmentation in Colonoscopic Frames Using CVC-ClinicDB Dataset

Comparative Analysis of U-Net and DeepLab for Automatic Polyp Segmentation in Colonoscopic Frames Using CVC-ClinicDB Dataset