Multimodal Objects Categorization by Fusing GMM and Multi-layer Perceptron

Published: Feb. 19, 2024

Multi-object recognition resolves the need to instantaneously identify numerous things, which is essential for developing an in-depth understanding of world vision. Variations in object features, occlusions, scale, perspective, and intricate backgrounds make this task challenging. The practical strategy categorizing objects into multiple categories using well-known benchmark datasets presented work. method extracts multi-object segmentation from difficult images advanced Gaussian mixture model (GMM) technique. A multi-layer perceptron (MLP) architecture that incorporates local descriptors detectors optimizes classification process. method's ability handle concerns like occlusion spatial complexity validated by extensive experiments a dataset, Corel 10k. advised overtakes cutting-edge techniques terms accuracy, with scores 87.40% accuracy 10k dataset.

Language: Английский

---

Efficient Vehicle Detection and Tracking using Blob Detection and Kernelized Filter

Published: Feb. 19, 2024

Smart conveyance techniques are crucial in modern society, enabling efficient traffic management and ensuring public safety. Therefore, this paper proposes a novel approach combining Blob detection the kernelled correlation filter (KCF)-based visual tracking method for vehicle detection, tracking, trajectory generation. Numerous preprocessing techniques, including gamma correction, contrast reduction, bilateral filtering, Mean Shift filtering applied to improve results of detection. Subsequently, KCF algorithm is employed robust The trajectories provide valuable information about flow vehicles, contributing better monitoring systems. suggested validated on publicly accessible KITTI benchmark dataset, demonstrate that integration blob KCF-based along with significantly improves accuracy. Our achieved an accuracy S2% 86% vehicles. There lot potential research improving safety, management, technique yields useful regarding trajectories, which enables thorough study patterns.

Language: Английский

---

A Hybrid Approach for Sports Activity Recognition Using Key Body Descriptors and Hybrid Deep Learning Classifier

Sensors, Journal Year: 2025, Volume and Issue: 25(2), P. 441 - 441

Published: Jan. 13, 2025

This paper presents an approach for event recognition in sequential images using human body part features and their surrounding context. Key points were approximated to track monitor presence complex scenarios. Various feature descriptors, including MSER (Maximally Stable Extremal Regions), SURF (Speeded-Up Robust Features), distance transform, DOF (Degrees of Freedom), applied skeleton points, while BRIEF (Binary Independent Elementary HOG (Histogram Oriented Gradients), FAST (Features from Accelerated Segment Test), Optical Flow used on silhouettes or full-body capture both geometric motion-based features. Feature fusion was employed enhance the discriminative power extracted data physical parameters calculated by different extraction techniques. The system utilized a hybrid CNN (Convolutional Neural Network) + RNN (Recurrent classifier recognition, with Grey Wolf Optimization (GWO) selection. Experimental results showed significant accuracy, achieving 98.5% UCF-101 dataset 99.2% YouTube dataset. Compared state-of-the-art methods, our achieved better performance recognition.

Language: Английский

---

Smart Traffic Monitoring through Drone Images via Yolov5 and Kalman Filter

Published: Feb. 19, 2024

Monitoring traffic is of vital relevance in our modern scenario. Over the years, standard ways capturing data, such as induction loops and camcorders, were applied for this aim. Nevertheless, introduction unmanned aerial vehicles (UAVs) has opened new possibilities subject, leading to extensive study computer vision. However, there are still issues object detection tracking because complexities that come from a high number objects, shifting heights (UAVs), variable lighting conditions. Our provides approach combines centroid with YOLOv5 algorithm successfully monitor identify vehicles. To acquire an accurate analysis, we began method by aligning referencing images. These basic measurements assisted advancement next phases covering feature extraction, segmentation region interest, tasks connected tracking. The effectiveness technique was tested applying it VAID dataset, resulting amazing levels accuracy demonstrated recommended solution. system exhibited unparalleled 98.5% reasonable 90.9% tests. validation indicates robustness efficacy designed solving obstacles given sophisticated surveillance

Language: Английский

---

Dynamic Adoptive Gaussian Mixture Model for Multi-Object Detection Over Natural Scenes

Published: Feb. 19, 2024

Recent strides in visionary technologies have profoundly impacted the domain of multi-object recognition, serving as a linchpin transformative applications such augmented reality integration, robotic navigation, and autonomous driving. This research introduces an innovative paradigm for detection, marked by substantial advancements accuracy efficiency. The proposed methodology combines Dynamic Adoptive Gaussian Mixture Model (DAGMM) with advanced feature extraction fusion techniques. cutting-edge Specifically, we leverage capabilities HOG, Akaze Brisk methods to capture intricate object descriptors rich data profile. Subsequently, introduce nuanced approach involving weighted variance thresholds fusion, enhancing discriminative prowess extracted features. Finally, classification, incorporate Convolutional Neural Networks (CNN). We meticulously assessed performance our model two demanding datasets: UIUC sports Caltech-101. Our DAGMM showcased exceptional proficiency, delivering impressive segmentation rates $86.9 \%$ $\mathbf{8 1. 1 \%}$ over Caltech-101, respectively, when evaluated against rigorously self-annotated ground truth. Furthermore, achieved $87.2 classification accuracy, validating its effectiveness.

Language: Английский

---

Robust Object Recognition with Genetic Algorithm and Composite Saliency Map

Published: Feb. 19, 2024

Object detection and recognition have emerged as crucial research areas, driven by the dire need to address challenges posed identifying recognizing objects in diverse domains. Accurate object identification is required for tasks such autonomous navigation, video surveillance, precision agriculture, medical imaging. We propose a method based on unique combination of machine learning techniques recognition. The proposed approach starts applying K-means segmentation input image cluster similar regions colors. Next, composite saliency map generated output. Subsequently, technique extracting from accomplished through employment connected pixel extraction method. Ultimately, Genetic Algorithm utilized optimize decision tree classifier purpose MSRC dataset was assess approach, resulting an accuracy 81.5%, accompanied 83.3% recall 86.1%. results depicts that our highly valuable accurate categorization, wherein utilization significantly enhances classification.

Language: Английский

---

Data Driven Approach to Leaf Recognition: Logistic Regression for Smart Agriculture

Published: Feb. 19, 2024

Artificial Intelligence has been involved in the restructuring of farming practices. Compared to conventional methods that using pesticides remove weeds, smart and precision agriculture increased productivity produced healthy crops. Plant leaves with varying behaviors should be separated as one measures preserving yield quality. We suggest logistic regression implement a four-step AI-based process classify different leaves. Pre-processing, segmentation feature extraction, have used for images classification. Various grapevine swedish work by applying suggested technique two datasets. image size $256\times 256$. During pre-processing, median filter was used, k-means segmentation, Kaze Blob were selected extraction. Logistic Our findings demonstrate strategy beats alternative approaches, yielding an accuracy 88% on leave dataset 83% datasets pertaining

Language: Английский

---

An Efficient Approach to Sports Rehabilitation and Outcome Prediction Using RNN-LSTM

Mobile Networks and Applications, Journal Year: 2024, Volume and Issue: unknown

Published: June 12, 2024

Language: Английский

---

CNN-Based Object Detection via Segmentation Capabilities in Outdoor Natural Scenes

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 84984 - 85000

Published: Jan. 1, 2024

Language: Английский

---

A wearable-based sports health monitoring system using CNN and LSTM with self-attentions

PLoS ONE, Journal Year: 2023, Volume and Issue: 18(10), P. e0292012 - e0292012

Published: Oct. 11, 2023

Sports performance and health monitoring are essential for athletes to maintain peak avoid potential injuries. In this paper, we propose a sports system that utilizes wearable devices, cloud computing, deep learning monitor the status of persons. The consists device collects various physiological parameters server contains model predict sportsperson's status. proposed combines Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), self-attention mechanisms. is trained on large dataset persons' data achieves an accuracy 93%, specificity 94%, precision 95%, F1 score 92%. person can access using their mobile phone receive report status, which be used make any necessary adjustments training or competition schedule.

Language: Английский

---