Object Detection Research Articles

Brain tumor detection using a MobileNetV2-SSD model with modified feature pyramid network levels

Brain tumors, a subset of these malignancies, demand accurate and efficient diagnosis. Traditional methods use non-invasive medical imaging like magnetic resonance imaging (MRI) and computed tomography (CT). Although necessary for diagnosis, manual brain MRI picture segmentation is tedious and time-consuming. Using deep learning is a promising solution. This study proposes an innovative approach for brain tumor detection, focusing on meningioma tumors. Utilizing threshold-based segmentation, the MobileNetV2 architecture, a modified feature pyramid network (FPN), and single shot MultiBox detector (SSD), our model achieves precise localization and object detection. Pre-processing techniques such as grayscale conversion, histogram equalization, and Gaussian filtering enhance the MRI image quality. Morphological operations and thresholding facilitate tumor segmentation. Data augmentation and a meticulous dataset division aid in model generalization. The architecture combines MobileNetV2 as a feature extractor, SSD for object detection, and FPN for detecting small objects. Modifications, including lowering the minimum FPN level, enhance small object detection accuracy. The proposed model achieved a recall value of around 98% and a precision value of around 89%. Additionally, the proposed model achieved approximately 93% on both the dice similarity coefficient (DSC) value and the index of similarity. Based on the promising results, our research holds significant advancements for the field of medical imaging and tumor detection.

An efficient object detection by autonomous vehicle using deep learning

The automation industries have been developing since the first demonstration in the period 1980 to 2000 it is mainly used on automated driving vehicle. Now a day’s automotive companies, technology companies, government bodies, research institutions and academia, investors and venture capitalists are interested in autonomous vehicles. In this work, object detection on road is proposed, which uses deep learning (DL) algorithms. You only look once (YOLO V3, V4, V5). In this system object detection on the road data set is taken as input and the objects are mainly on-road vehicles, traffic signals, cars, trucks and buses. These inputs are given to the models to predict and detect the objects. The Performance of the proposed system is compared with performance of deep learning algorithms convolution neural network (CNN). The proposed system accuracy greater than 76.5% to 93.3%, mean average precision (Map) and frame per second (FPS) are 0.895 and 43.95%.

Swin transformer adaptation into YOLOv7 for road damage detection

Highways are an important component of any country. However, some highways in Indonesia endanger users while maintaining road safety. Crack detection early in the deterioration process can prevent further damage and lower maintenance costs. A recent study sought to develop a method for detecting road damage by combining the road damage detection (RDD) dataset with generative adversarial network technology and data augmentation to improve training. The current study aims to broaden the you only look once (YOLO) framework by incorporating the Swin Transformer into the chiral stationary phases (CSP) component of YOLOv7, with the goal of improving object detection accuracy in a variety of visual scenarios. The study compares the performance of various object detection models with varying parameters and configurations, such as YOLOv5l, YOLOv6l, YOLOv7-tiny, YOLOv7, and YOLOv7x. YOLOv5l has 46 million parameters and 108 billion floating point operations per second (FLOPS), whereas YOLOv6l has 59.5 million parameters and 150 billion FLOPS. With 31 million parameters and 140 billion FLOPS, the YOLOv7-swin model performs best with mean average precision (mAP), mAP_0.50 of 0.47. and mAP_0.5:0.95 of 0.232. The experimental results show that our YOLOv7-swin model outperforms both YOLOv7x and YOLOv7-tiny. The proposed model significantly improves object detection accuracy while keeping complexity and performance in balance.

Citrus leaf disease detection through deep learning approach

The majority of people in the world directly or indirectly depend on agriculture. Plant diseases are a significant threat to agricultural production and food security. Due to its high nutritional value, citrus fruit is one of the most abundant fruits in the world. However, different diseases are responsible for degraded citrus production as well as financial losses to the farmers. Traditionally, visual observation by experts has been attended to diagnose plant diseases. Usually, plant leaf disease recognition methods mainly rely on expert experiences to manually extract the colour, composition, and other features of diseased leaf images. Black spot, greening, canker, and melanoses are four common citrus leaf diseases. Rapid and accurate diagnosis of these diseases is a demand of time. Deep learning is a promising solution to these problems. There are different types of deep learning architecture like ImageNet, GoogleNet, VGG16, ResNet50, and InceptionV3, which show promising results in different object detection. Though most of these benchmark models give almost similar accuracy. However, this paper uses two deep learning models to find the better ones for the detection of citrus leaf disease detection. Hence, InceptionV3 outperforms VGG16 in terms of accuracy.

Object-based image retrieval and detection for surveillance video

With technological advancement worldwide, the video surveillance market is growing drastically in a versatile field. Monitoring, browsing, and retrieving a specific object in a long video becomes difficult due to the enormous amount of data produced by the surveillance camera. With limitations on human resources and browsing time, there is a need for a new video analytics model to handle more complex tasks, such as object detection and query retrieval. The current approach involves techniques like unsupervised segmentation, multiscale segmentation, and feature-based descriptions. However, these methods often encounter extensive space and time consumption challenges. A solution has been developed for retrieving targeted objects from surveillance videos via user queries, employing a graphical interface for input. Extracting relevant frames based on user-entered text queries is enabled through using YOLOv8 for object detection. Users interact through a graphical user interface deployed on a Jetson Xavier Development board. The system's outcome is a time-efficient and highly accurate automated model for object detection and query retrieval, eliminating human errors associated with manually locating objects in videos upon user queries.

YOLO-based object detection performance evaluation for automatic target aimbot in first-person shooter games

First-person shooter (FPS) focuses on first-person perspective action gameplay, with gunfights usually giving the player a choice of weapons, significantly impacting how the player approaches or strategies. General military-themed FPS games have realistic models with actual weapons’ shapes and characteristics. This type of game requires high aiming accuracy while using a mouse on a PC. However, not all players have a fast response time in knowing the surrounding situation. New players may need aid when targeting enemies in the FPS world. One popular yet underhanded method is injecting a program code using a dynamic-link library (DLL) to manipulate memory and asset data from the game. Instead of DLL, we promote a novel approach using the player’s real-time game screen, detecting the person without injecting program code into the game. The you only look once (YOLO) algorithm is used as an object detector model since it can process images in real time for up to 45 frames per second. The proposed object detection has an outstanding performance with 65% accuracy, 98% precision, and 61% recall of 51 tests for each game. YOLO’s fastest detection speed produces an average of 35 FPS on the YOLO tiny variant using a mixed precision (half) graphics processing unit (GPU).

A dataset for computer-vision-based fig fruit detection in the wild with benchmarking you only look once model detector

The image datasets that are most widely used for training deep learning models are specifically developed for applications. This study introduces a novel dataset aimed at augmenting the existing data for the identification of figs in their natural habitats, specifically in the wilderness. In the present study, researchers have generated numerous image datasets specifically for object detection focus on applications in agriculture. Regrettably, it is exceedingly difficult for us to obtain a specialized dataset specifically designed for detecting figs. To tackle this issue, a grand total of 462 photographs of fig fruits were gathered. The augmentation technique was utilized to substantially increase the size of the dataset. Ultimately, we conduct an examination of the dataset by doing a baseline performance study for bounding-box detection using established object detection methods, specifically you only look once (YOLO) version 3 and YOLOv4. The performance obtained on the test photos of our dataset is satisfactory. For farmers, the capacity to identify and oversee fig fruits in their natural or developed environments can be highly advantageous. The detecting device offers instantaneous data regarding the quantity of mature figs, facilitating decision-making procedures.

A two-stage approach for aircraft detection with convolutional neural network

Over the past few years, object detection has experienced remarkable advancements, primarily attributable to significant progress in deep learning architectures. Nonetheless, the task of identifying aircraft targets within remote sensing images remains a challenging and actively explored area. Presently, there are two main approaches employed for this task: one utilizing convolutional neural network (CNN) techniques and the other relying on conventional methods. In this work, a CNN based architecture is proposed to recognize aircraft types using remote sensing images. The experiments performed on multi-type aircraft remote sensing images (MTARSI) dataset show that the proposed architecture achieves 97.07%, 94.81%, and 94.44% accuracy rates for training, validation and testing sets. The results approve that, the architecture outperforms state of the art models.

Efficient traffic signal detection with tiny YOLOv4: enhancing road safety through computer vision

As decades go by, technology advances and everything around us becomes smarter, such as televisions, mobile phones, robots, and so on. Artificial intelligence (AI) is applied in these technologies where AI assists the computer in making judgments like humans, and this intelligence is artificially fed to the model. The self-driving technique is a developing technology. Autonomous driving has been a broad and fast-expanding technology over the last decade. This model is carried out using the tiny you only look once (YOLO) algorithm. YOLO is mainly used for object detection classification. Tiny YOLO model is explored for the traffic signal detection. ROBI FLOW dataset is used for object detection which contains 2000+ image data to train the tiny YOLO model for traffic signal detection in real time. This model gives an improved accuracy and lightweight implementation compared to other models. Tiny YOLO is fast and accurate model for real-time traffic signal detection.

CycleInSight: An enhanced YOLO approach for vulnerable cyclist detection in urban environments

As urbanization continues to reshape transportation, the safety of cyclists in complex traffic environments has become a pressing concern. In response to this challenge, our research introduces a CycleInSight framework, which harnesses advanced deep learning and computer vision techniques to enable precise and efficient cyclist detection in diverse urban settings. Utilizing you only look once version 8 (YOLOv8) object detection algorithm, the proposed model aims to detect and localize vulnerable cyclists near vehicles equipped with onboard cameras. Our research presents comprehensive experimental results demonstrating its effectiveness in identifying vulnerable cyclists amidst dynamic and challenging traffic conditions. With an impressive average precision of 90.91%, our approach outperforms existing models while maintaining efficient inference speeds. By effectively identifying and tracking cyclists, this framework holds significant potential to enhance urban traffic safety, inform data-driven infrastructure planning, and support the development of advanced driver assistance systems and autonomous vehicles.

Effectiveness of VGG19 in deep learning for brain tumor detection

Image processing in the diagnosis of disease is one of the jobs that is currently developing in the world of health. Diagnosis is carried out by utilizing the role of image processing to provide a level of accuracy in diagnosis results and provide efficiency to medical personnel. This research aims to develop a brain tumor object detection process using a deep learning (DL) approach to magnetic resonance images (MRI) images. This development was carried out to optimize the brain tumor diagnosis process by playing the role of the image extraction process. This research dataset was sourced from the M. Djamil Padang Provincial General Hospital with a total of 3370 MRI images. The results of this work report show that DL performance is capable of carrying out the detection process automatically with an accuracy level of 97,83%. The results of the development of the extraction process can work effectively in ensuring brain tumor objects are precise and accurate. Overall, this research can make a major contribution to maximizing the diagnosis process and assisting medical personnel in the early treatment of brain tumor patients.

Railroad Catenary Insulator Fault Detection Based on Improved Faster R-CNN

Background:: The railroad catenary insulator, which is a crucial component of the catenary system and is situated between the pillar and wrist arm, is crucial for electrical conductor isolation, electrical equipment insulation, mechanical load bearing, anti-fouling, and anti-leakage. The catenary insulators will experience tarnished flash, breakage, insulation strength deterioration, and other issues as a result of the long-term outside unfavorable working circumstances. The train electrical system's ability to operate normally is greatly hampered by these problems. Although there are many patents and articles related to insulator fault detection, the precision is not high enough. Therefore, it is crucial to improve the precision of catenary insulator fault detection. Objective:: An improved region-based convolutional neural networks (Faster R-CNN)-based fault detection method for railway catenary insulators is proposed in response to the long detection time of the conventional railroad catenary insulator fault, the low precision of the catenary insulator fault detection for occlusion and truncation, the poor performance of multi-scale object detection, and the processing of class unbalance problem. Methods:: The Faster R-CNN is optimized from four perspectives: feature extraction, feature fusion, candidate box screening, and loss function, in accordance with the properties of the catenary insulator. First, to solve the problem of multi-scale catenary insulator fault detection, convolutional block attention module (CBAM) and feature pyramid network (FPN) are used to fuse the deep feature and shallow features of the image. This results in a feature map with more critical semantic information and higher resolution. After that, the weighted non-maximum suppression (WNMS) algorithm improved by distance-intersection over union (DIOU) and Gaussian weighting function is used instead of the traditional NMS algorithm, which effectively introduces the overlap of detection frames into the confidence level and makes full use of the effective information of the detection frames. Finally, the improved Focal loss is used as the classification loss, and the focusing parameter and the balance factor of the Focal Loss are adjusted dynamically to solve the problem of sample imbalance and difficult sample identification in the model better. Results:: The effects of SSD, YOLOV3, traditional Faster R-CNN and improved Faster R-CNN models are tested on the contact network insulator fault detection dataset constructed in this paper, and the experimental results show that the improved Faster R-CNN has higher precision, recall, and mAP compared to the other detection models, which reach 94.31%, 96.68% and 95.22%, respectively. Conclusion:: The results of the experiments demonstrate that this method may successfully detect the faults in different scale catenary insulators. It can effectively detect truncated, obscured faulty catenary insulators. It has higher precision and recall and provides a reliable reference for maintaining faulty insulators in railway catenary.

A deep learning-based method for estimating the main stem length of sweet potato seedlings

The length of the main stem of sweet potato seedlings is a crucial criterion for determining whether to harvest it. However, accurately estimating the main stem length is challenging due to two growth forms: leaf occlusion and self-curvature, which complicate automated harvesting and grading. This paper considers the morphological characteristics of sweet potato seedlings and proposes a main stem reconstruction and length estimation algorithm based on a two-stage classifier and feature extraction and search. Semantic segmentation is used to obtain masks of the stems. This is followed by image processing to segment clustered skeletons into singular lines, facilitating feature analysis of the main stem. Object detection is employed to identify regions of interest such as roots, growth joints, and stem-leaf intersections. Using the category and centroid information obtained from object detection, a prioritized search relationship is established among the line segments extracted through image processing. Subsequently, the main stems are classified and reconstructed based on constraints including direction, distance, and curvature. Finally, utilizing stereo depth information, the main stem is divided into four parts to estimate its length. Experimental results demonstrate that the algorithm achieves effective main stem reconstruction under four conditions: un-occluded, partially occluded, curved main stem, and low-density clustered multi-seedling states. The algorithm achieves a median length estimation error of 1.7 cm, with maximum errors of 3.2 cm and minimum errors of 0.3 cm. It provides an automated solution for length estimation during sweet potato seedling harvesting and grading.The length of sweet potato seedling main stems is critical for harvest decisions, but accurately measuring it is challenging due to leaf occlusion and self-curvature. This study proposes a method for stem reconstruction and length estimation using a two-stage classifier, feature extraction, and semantic segmentation to obtain stem masks. Image processing separates clustered skeletons into singular lines for detailed analysis. Object detection identifies roots, stem-leaf intersections, and establishes a search relationship among segments. Stems are reconstructed based on directional, distance, and curvature constraints, then divided into four parts for length estimation using stereo depth information. Experimental results demonstrate effectiveness under various conditions: un-occluded, partially occluded, curved stems, and low-density clusters. Median length estimation error is 1.7 cm, with maximum and minimum errors of 3.2 cm and 0.3 cm respectively. This automated method provides a reliable solution for sweet potato seedling harvest and grading.

HDConv: Heterogeneous kernel-based dilated convolutions

Dilated convolution has been widely used in various computer vision tasks due to its ability to expand the receptive field while maintaining the resolution of feature maps. However, the critical challenge is the gridding problem caused by the isomorphic structure of the dilated convolution, where the holes filled in the dilated convolution destroy the integrity of the extracted information and cut off the relevance of neighboring pixels. In this work, a novel heterogeneous dilated convolution, called HDConv, is proposed to address this issue by setting independent dilation rates on grouped channels while keeping the general convolution operation. The heterogeneous structure can effectively avoid the gridding problem while introducing multi-scale kernels in the filters. Based on the heterogeneous structure of the proposed HDConv, we also explore the benefit of large receptive fields to feature extraction by comparing different combinations of dilated rates. Finally, a series of experiments are conducted to verify the effectiveness of some computer vision tasks, such as image segmentation and object detection. The results show the proposed HDConv can achieve a competitive performance on ADE20K, Cityscapes, COCO-Stuff10k, COCO, and a medical image dataset UESTC-COVID-19. The proposed module can readily replace conventional convolutions in existing convolutional neural networks (i.e., plug-and-play), and it is promising to further extend dilated convolution to wider scenarios in the field of image segmentation.

A novel system for the classification of zinc-plated components by benchmarking deep neural networks

Artificial intelligence (AI) has become the new technique for solving the most complicated challenges related to big data, image recognition, object detection, and prediction issues due to the fourth industrial revolution. This article introduces the development and implementation of a zinc-plated component recognition system within a manufacturing process using deep learning (DL) techniques. This paper aims to train and evaluate different DL algorithms to recognize five zinc-plated components and one assembly tray under different ambient light conditions and finishing. The proposed method begins with creating a custom dataset of 5.5K images, with six classes that match the assembly components. However, before the training stage starts, a subset between 5% and 10% is created. Eight deep neural networks have been benchmarked: Swing Transformer, Convnextv2, Xception41, Data-efficient image Transformer, Inception_v4, ViT, Resnet50 and Efficientnet_b0. The training stage is performed with the k-fold method using k = 6 and a progressive data size between different runs with an average of data utilized during training of 8%. In this study, convnextv2_tiny exhibits the highest F1-Score with 99.177%, followed by swin_tiny_patch4_window7_224 with 99.175%, while resnet50 achieved the lowest F1-Score at 95.886%. Therefore, the new generation of convolutional architectures is still better for classifying images. Thus, these algorithms can be adopted as good classifier models for zinc-plated component classification in the industrial manufacturing environment.

Improved YOLOv5 foreign object detection for transmission lines

Improved YOLOv5 foreign object detection for transmission lines

AFO-SLAM: an improved visual SLAM in dynamic scenes using acceleration of feature extraction and object detection

Abstract In visual simultaneous localization and mapping (SLAM) systems, traditional methods often excel due to rigid environmental assumptions, but face challenges in dynamic environments. To address this, learning-based approaches have been introduced, but their expensive computing costs hinder real-time performance, especially on embedded mobile platforms. In this article, we propose a robust and real-time visual SLAM method towards dynamic environments using acceleration of feature extraction and object detection (AFO-SLAM). First, AFO-SLAM employs an independent object detection thread that utilizes YOLOv5 to extract semantic information and identify the bounding boxes of moving objects. To preserve the background points within these boxes, depth information is utilized to segment target foreground and background with only a single frame, with the points of the foreground area considered as dynamic points and then rejected. To optimize performance, CUDA program accelerates feature extraction preceding point removal. Finally, extensive evaluations are performed on both TUM RGB-D dataset and real scenes using a low-power embedded platform. Experimental results demonstrate that AFO-SLAM offers a balance between accuracy and real-time performance on embedded platforms, and enables the generation of dense point cloud maps in dynamic scenarios.

Weak-Texture Seafloor and Land Image Matching Using Homography-Based Motion Statistics with Epipolar Geometry

The matching of remote sensing images is a critical and necessary procedure that directly impacts the correctness and accuracy of underwater topography, change detection, digital elevation model (DEM) generation, and object detection. The texture of images becomes weaker with increasing water depth, and this results in matching-extraction failure. To address this issue, a novel method, homography-based motion statistics with an epipolar constraint (HMSEC), is proposed to improve the number, reliability, and robustness of matching points for weak-textured seafloor images. In the matching process of HMSEC, a large number of reliable matching points can be identified from the preliminary matching points based on the motion smoothness assumption and motion statistics. Homography and epipolar geometry are also used to estimate the scale and rotation influences of each matching point in image pairs. The results show that the matching-point numbers for the seafloor and land regions can be significantly improved. In this study, we evaluated this method for the areas of Zhaoshu Island, Ganquan Island, and Lingyang Reef and compared the results to those of the grid-based motion statistics (GMS) method. The increment of matching points reached 2672, 2767, and 1346, respectively. In addition, the seafloor matching points had a wider distribution and reached greater water depths of −11.66, −14.06, and −9.61 m. These results indicate that the proposed method could significantly improve the number and reliability of matching points for seafloor images.

SMART BABY MONITORING SYSTEM USING YOLO V8 ALGORITHM

The Smart Baby Monitoring System using the YOLO V8 algorithm is designed to enhance infant monitoring by leveraging advanced computer vision techniques. This project utilizes YOLO (You Only Look Once) version 8, a state-of-the-art object detection algorithm, implemented with Python and frameworks like Tensor Flow or PyTorch, to detect and track objects in real-time video feeds. The system incorporates features for facial recognition to identify known caregivers and alert mechanisms for unusual activities or emergencies. The user interface provides real-time alerts, visualizations, and historical data analysis for caregivers via a web or mobile application. By leveraging YOLO V8's efficiency in object detection and Python's capabilities for data processing and integration, this system aims to enhance safety, improve care giving efficiency, and provide peace of mind to parents and caregivers.

Lane and Object Detection using YOLO: Indian Roads Scenario

This research explores the feasibility of You Only Look Once (YOLO), a deep learning object detection algorithm, for lane and object detection in challenging Indian road environments. Traditional methods struggle with faded lane markings, dense and diverse traffic, and unpredictable scenarios. YOLO's speed and accuracy make it suitable for real-time ADAS applications. The methodology leverages Roboflow, a platform for computer vision tasks, to explore data acquisition, model selection, training, and evaluation. This research aims to contribute to developing safer and more reliable ADAS systems for Indian roads.