object detection

Impact of Architectural Scaling in YOLO11 Models on Object Detection Accuracy and Inference Performance in UAV Systems

This paper investigates the impact of architectural scaling in YOLO-family neural object detectors on object detection performance in unmanned aerial vehicle (UAV) systems under CPU-only inference conditions without hardware acceleration. Standard nano and small model configurations are analyzed, along with an intermediate model obtained through controlled width scaling of the network. Experimental evaluation is conducted on an embedded Raspberry Pi 5 platform under fixed hardware and software conditions using ONNX Runtime, ensuring fair comparability of the models.

Optimization Method for Reducing the Size and Delay of Deep Learning Models in the Military Vehicle Recognition System

This paper presents a method for optimizing deep learning models used in real-time military equipment recognition systems. A key contribution of the work is the application of exponential data augmentation, which significantly increases dataset diversity without requiring additional real-world data. The augmented dataset has been used to train a YOLO-based object detection model capable of recognizing various types of military vehicles, including tanks, armored personnel carriers, and infantry fighting vehicles.

Underwater Trash Detection with YOLO: Enhancing Environmental Monitoring in Aquatic Ecosystems

This study explores the application of YOLOv8, a cutting-edge object detection framework, to detect underwater waste.  As concerns about protecting marine ecosystems grow, the effective identification and removal of underwater debris have become critical.  The research adapts YOLOv8 to address challenges such as variable lighting, distortions, and occlusions in underwater environments.  A specialized dataset, containing various underwater landscapes and debris types, was created to train and evaluate the model.  Experimental results show strong performance across differ

Exponential Data Augmentation Methods for Improving Yolo Performance in Computer Vision Tasks

The article examines data augmentation methods in the task of image recognition, specifically introducing the exponential augmentation approach to enhance the performance of deep neural networks, particularly YOLO, in object detection tasks. The proposed methodology is based on the sequential and repeated application of various transformations, including horizontal and vertical flipping, 90° rotation, Gaussian Blur, brightness and contrast adjustment.

Advanced YOLO models for real-time detection of tomato leaf diseases

The increasing focus on smart agriculture in the last decade can be attributed to various factors, including the adverse effects of climate change, frequent extreme weather events, increasing population, the necessity for food security, and the scarcity of natural resources.  The government of Morocco adopts preventative measures to combat plant illnesses, specifically focusing on tomatoes.  Tomatoes are widely acknowledged as one of the most important vegetable crops, but they are highly vulnerable to several diseases that significantly decrease their productivity.  De

Intelligent driver assistance systems based on computer vision and deep learning

This article presents an integrated Advanced Driver Assistance System (ADAS) that combines several key functional modules, such as collision warning, lane detection, traffic sign recognition, and pothole detection, which are implemented using modern deep learning models, particularly YOLOv8n. The system is optimized for devices with limited computational resources, such as Raspberry Pi or NVIDIA Jetson Nano, by employing a modular architecture and parallel data processing to ensure realtime performance.

USING NEURAL NETWORKS TO IDENTIFY OBJECTS IN AN IMAGE

A modified neural network model based on Yolo V5 was developed and the quality metrics of object classification on video images built on the basis of existing known basic neural network architectures were compared. The application of convolutional neural networks for processing images from video surveillance cameras is considered in order to develop an optimized algorithm for detecting and classifying objects on video images. The existing models and architectures of neural networks for image analysis were analyzed and compared.

PREVENTING POTENTIAL ROBBERY CRIMES USING DEEP LEARNING ALGORITHM OF DATA PROCESSING

Recently, deep learning technologies, namely Neural Networks [1], are attracting more and more attention from businesses and the scientific community, as they help optimize processes and find real solutions to problems much more efficiently and economically than many other approaches. In particular, Neural Networks are well suited for situations when you need to detect objects or look for similar patterns in videos and images, making them relevant in the field of information and measurement technologies in mechatronics and robotics.

Improving pedestrian segmentation using region proposal-based CNN semantic segmentation

Pedestrian segmentation is a critical task in computer vision, but it can be challenging for segmentation models to accurately classify pedestrians in images with challenging backgrounds and luminosity changes, as well as occlusions.  This challenge is further compounded for compressed models that were designed to deal with the high computational demands of deep neural networks.  To address these challenges, we propose a novel approach that integrates a region proposal-based framework into the segmentation process.  To evaluate the performance of the proposed framework,

Analysis of Algorithms for Searching Objects in Images Using Convolutional Neural Network

The problem of finding objects in images using modern computer vision algorithms has been considered. The description of the main types of algorithms and methods for finding objects based on the use of convolutional neural networks has been given. A comparative analysis and modeling of neural network algorithms to solve the problem of finding objects in images has been conducted. The results of testing neural network models with different architectures on data sets VOC2012 and COCO have been presented.