Real-time Object Detection Research Articles

MSQuant: Efficient Post-Training Quantization for Object Detection via Migration Scale Search

YOLO (You Only Look Once) has become the dominant paradigm in real-time object detection. However, deploying real-time object detectors on resource-constrained platforms faces challenges due to high computational and memory demands. Quantization addresses this by compressing and accelerating CNN models through the representation of weights and activations with low-precision values. Nevertheless, the quantization difficulty between weights and activations is often imbalanced. In this work, we propose MSQuant, an efficient post-training quantization (PTQ) method for CNN-based object detectors, which balances the quantization difficulty between activations and weights through migration scale. MSQuant introduces the concept of migration scales to mitigate this disparity, thereby improving overall model accuracy. An alternating search method is employed to optimize the migration scales, avoiding local optima and reducing quantization error. We select YOLOv5 and YOLOv8 models as the PTQ baseline, followed by extensive experiments on the PASCAL VOC, COCO, and DOTA datasets to explore various combinations of quantization methods. The results demonstrate the effectiveness and robustness of MSQuant. Our approach consistently outperforms other methods, showing significant improvements in quantization performance and model accuracy.

Crop Identification by using Real Time Object Detection

Crop Identification by using Real Time Object Detection

Hybrid CNN-Transformer Model for Accurate Impacted Tooth Detection in Panoramic Radiographs

Background/Objectives: The integration of digital imaging technologies in dentistry has revolutionized diagnostic and treatment practices, with panoramic radiographs playing a crucial role in detecting impacted teeth. Manual interpretation of these images is time consuming and error prone, highlighting the need for automated, accurate solutions. This study proposes an artificial intelligence (AI)-based model for detecting impacted teeth in panoramic radiographs, aiming to enhance accuracy and reliability. Methods: The proposed model combines YOLO (You Only Look Once) and RT-DETR (Real-Time Detection Transformer) models to leverage their strengths in real-time object detection and learning long-range dependencies, respectively. The integration is further optimized with the Weighted Boxes Fusion (WBF) algorithm, where WBF parameters are tuned using Bayesian optimization. A dataset of 407 labeled panoramic radiographs was used to evaluate the model’s performance. Results: The model achieved a mean average precision (mAP) of 98.3% and an F1 score of 96%, significantly outperforming individual models and other combinations. The results were expressed through key performance metrics, such as mAP and F1 scores, which highlight the model’s balance between precision and recall. Visual and numerical analyses demonstrated superior performance, with enhanced sensitivity and minimized false positive rates. Conclusions: This study presents a scalable and reliable AI-based solution for detecting impacted teeth in panoramic radiographs, offering substantial improvements in diagnostic accuracy and efficiency. The proposed model has potential for widespread application in clinical dentistry, reducing manual workload and error rates. Future research will focus on expanding the dataset and further refining the model’s generalizability.

A small underwater object detection model with enhanced feature extraction and fusion

In the underwater domain, small object detection plays a crucial role in the protection, management, and monitoring of the environment and marine life. Advancements in deep learning have led to the development of many efficient detection techniques. However, the complexity of the underwater environment, limited information available from small objects, and constrained computational resources make small object detection challenging. To tackle these challenges, this paper presents an efficient deep convolutional network model. First, a CSP for small object and lightweight (CSPSL) module is introduced to enhance feature retention and preserve essential details. Next, a variable kernel convolution (VKConv) is proposed to dynamically adjust the convolution kernel size, enabling better multi-scale feature extraction. Finally, a spatial pyramid pooling for multi-scale (SPPFMS) method is presented to preserve the features of small objects more effectively. Ablation experiments on the UDD dataset demonstrate the effectiveness of the proposed methods. Comparative experiments on the UDD and DUO datasets demonstrate that the proposed model delivers the best performance in terms of computational cost and detection accuracy, outperforming state-of-the-art methods in real-time underwater small object detection tasks.

Camouflaged Object Detection System

The detection of camouflaged objects is crucial for applications in surveillance, wildlife monitoring, and military scenarios, where objects blend seamlessly into their surroundings. This review consolidates 15 influential research studies covering advancements in datasets, models, real-time detection technologies, and multimodal approaches. The focus is on implementing YOLOv8, a state-of-the-art real-time object detection model, using the ACD1K dataset, which is specifically designed for military surveillance. By synthesizing methodologies, evaluation metrics, and applications, this paper highlights significant progress in camouflaged object detection (COD) and identifies ongoing challenges in computational efficiency, dataset diversity, and real-world adaptability. Key Words: Camouflaged object detection, YOLOv8, ACD1K dataset, military surveillance, real-time detection.

SST-YOLOv5s: advancing real-time blood cell object detection through multi-headed attention mechanism

SST-YOLOv5s: advancing real-time blood cell object detection through multi-headed attention mechanism

Model compression for real-time object detection using rigorous gradation pruning.

Achieving lightweight real-time object detection necessitates balancing model compression with detection accuracy, a difficulty exacerbated by low redundancy and uneven contributions from convolutional layers. As an alternative to traditional methods, we propose Rigorous Gradation Pruning (RGP), which uses a desensitized first-order Taylor approximation to assess filter importance, enabling precise pruning of redundant kernels. This approach includes the iterative reassessment of layer significance to protect essential layers, ensuring effective detection performance. We applied RGP to YOLOv8 detectors and tested it on GTSDB, Seaships, and COCO datasets. On GTSDB, RGP achieved 80% compression of YOLOv8n with only a 0.11% drop in mAP0.5, while increasing frames per second (FPS) by 43.84%. For YOLOv8x, RGP achieved 90% compression, a 1.26% mAP0.5:0.95 increase, and a 112.66% FPS boost. Significant compression was also achieved on Seaships and COCO datasets, demonstrating RGP's robustness across diverse object detection tasks and its potential for advancing efficient, high-speed detection models.

Improving Performance of Real-Time Object Detection in Edge Device Through Concurrent Multi-Frame Processing

Improving Performance of Real-Time Object Detection in Edge Device Through Concurrent Multi-Frame Processing

Lightweight Network for Real-Time Object Detection in Fisheye Cameras

Lightweight Network for Real-Time Object Detection in Fisheye Cameras

영상 기반 드론 및 UAM 비상 착륙 알고리즘 개발

In this study, we developed an algorithm that utilizes drone footage to construct training data, applies deep learning techniques for object inference, identifies suitable landing zones, and designates these zones as emergency landing sites. In particular, we developed a drone-based algorithm using YOLO-Seg deep learning to identify emergency landing zones for drones and UAM vehicles. By incorporating diverse training data, the algorithm segments landing areas with an accuracy of 84%-98% under various conditions. For the training process, we employed the YOLO-Segmentation model, which is a real-time object detection algorithm, to achieve precise boundary segmentation for landing zones. This segmentation facilitates accurate area detection and validates the performance of the algorithm. Furthermore, training data captured at different times of the day were incorporated to account for various environmental factors, thereby enhancing the robustness of the algorithm. This versatile algorithm can be applied to a wide range of aerial mobility safety scenarios. The proposed algorithm has demonstrated application potential in diverse and extensive scenarios related to aerial mobility safety.

YOLOInsight: Artificial Intelligence-Powered Assistive Device for Visually Impaired Using Internet of Things and Real-Time Object Detection

YOLOInsight: Artificial Intelligence-Powered Assistive Device for Visually Impaired Using Internet of Things and Real-Time Object Detection

Robotics Navigation Based on YOLO and Depth Camera

In recent years, significant advancements have been made in robotics, yet challenges remain in navigation, particularly for autonomous vehicles and bipedal robots. This paper provides a comprehensive review of three critical components in robotic navigation: YOLO neural networks, depth cameras, and A* path planning. Existing studies often address these aspects independently, lacking an integrated approach. Here, we systematically summarize and analyze the effectiveness of these techniques in navigation tasks. Through a literature review of recent publications, we compare and categorize various methods to assess trends in YOLO research and explore potential for integrated research in navigation. Furthermore, we analyze the strengths and limitations of each method in dynamic environments. The findings suggest that while YOLO and depth camera-based systems excel in real-time object detection and spatial awareness, they face challenges related to light sensitivity and high computational demands. Future research directions are proposed to enhance adaptability in complex environments, improve efficiency, and support cost-effective navigation solutions in robotics.

YH-RTYO: an end-to-end object detection method for crop growth anomaly detection in UAV scenarios

Background Small object detection via unmanned Aerial vehicle (UAV) is crucial for smart agriculture, enhancing yield and efficiency. Methods This study addresses the issue of missed detections in crowded environments by developing an efficient algorithm tailored for precise, real-time small object detection. The proposed Yield Health Robust Transformer-YOLO (YH-RTYO) model incorporates several key innovations to advance conventional convolutional models. The model features an efficient convolutional expansion module that captures additional feature information through extended branches while maintaining parameter efficiency by consolidating features into a single convolution during validation. It also includes a local feature pyramid module designed to suppress background interference during feature interaction. Furthermore, the loss function is optimized to accommodate various object scales in different scenes by adjusting the regression box size and incorporating angle factors. These enhancements collectively contribute to improved detection performance and address the limitations of traditional methods. Result Compared to YOLOv8-L, the YH-RTYO model achieves superior performance in all key accuracy metrics, with a 13% reduction in the scale of model. Experimental results demonstrate that the YH-RTYO model outperforms others in key detection metrics. The model reduces the number of parameters by 13%, facilitating deployment while maintaining accuracy. On the OilPalmUAV dataset, it achieves a 3.97% improvement in average precision (AP). Additionally, the model shows strong generalization on the RFRB dataset, with AP50 and AP values exceeding those of the YOLOv8 baseline by 3.8% and 2.7%, respectively.

Research on Object Detection Based on YOLO and Its Application in Tactile Paving Recognition

The YOLO model, as an advanced real-time object detection algorithm, has made significant progress in the field of object detection since its inception. It achieves fast and accurate object detection by transforming the object detection task into a regression problem and using a single neural network to predict multiple bounding boxes and category probabilities. The YOLO model has significant advantages in processing speed, detection accuracy, and real-time performance, and is widely used in fields such as autonomous driving, intelligent monitoring, and medical image analysis. This paper adopts a tactile paving detection method based on YOLOv8 and CBAM attention mechanism, and improves the model's ability to extract tactile paving features by introducing CBAM attention mechanism and loss function. This method shows good adaptability when dealing with the tactile paving environment under different lighting and weather conditions, can effectively identify diverse obstacles, and provides a strong guarantee for the travel safety of the visually impaired. In conclusion, YOLO model has broad prospects in object detection and tactile paving detection. The tactile paving detection method based on YOLOv8 and CBAM attention mechanism adopted in this paper provides new ideas and methods for the development of tactile paving detection technology, and has important practical application value. In the future, with the continuous optimization of performance and technological innovation of YOLO series models, further improvement of hardware performance, and the fusion of YOLO and other sensor data to form a multimodal detection system, YOLO models are expected to play an important role in more practical application scenarios.

Evaluating YOLO Models for Efficient Crack Detection in Concrete Structures Using Transfer Learning

The You Only Look Once (YOLO) network is considered highly suitable for real-time object detection tasks due to its characteristics, such as high speed, single-shot detection, global context awareness, scalability, and adaptability to real-world conditions. This work introduces a comprehensive analysis of various YOLO models for detecting cracks in concrete structures, aiming to assist in the selection of an optimal model for future detection and segmentation tasks. The YOLO models are initially trained on a dataset containing both images with and without cracks, producing a generalized model capable of extracting abstract features beneficial for crack detection. Subsequently, transfer learning is employed using a dataset that reflects real-world conditions, such as occlusions, varying crack sizes, and rotations, to further refine the model. Crack detection in concrete remains challenging due to the wide variation in crack sizes, aspect ratios, and complex backgrounds. To achieve optimal performance, we test different versions of YOLO, a state-of-the-art single-shot detector, and aim to balance inference speed and mean average precision (mAP). Our results indicate that YOLOv10 demonstrates superior performance, achieving a mean average precision (mAP) of 74.52% with an inference time of 19.5 milliseconds per image, making it the most effective among the models tested.

Implementation of Deep Learning Algorithm for Vehicle Count Monitoring System

Vehicle detection plays a crucial role in various applications such as traffic surveillance, license plate recognition, and the development of autonomous vehicles. The You Only Look Once (YOLO) object detection method is renowned for its high-speed real-time object detection capabilities. In this study, YOLO is employed to detect vehicles in images and videos. YOLO treats object detection as a direct regression problem for bounding boxes and class predictions. The aim of this research is to develop a vehicle counting system using the YOLO method. The Midpoint algorithm is utilized to calculate the midpoint between two points in a coordinate plane. Another objective is to analyze the strengths and weaknesses of the method and algorithm in the context of vehicle detection while identifying related research trends. The test results indicate that the system is capable of detecting vehicles with an average accuracy of 92.42% across four different time periods. In the morning, the system detected 156 vehicles (manual count: 147, accuracy: 94.23%); at midday, it detected 246 vehicles (manual count: 225, accuracy: 91.46%); in the evening, 377 vehicles were detected (manual count: 351, accuracy: 93.10%); and at night, the system identified 526 vehicles (manual count: 225, accuracy: 92.58%). This study contributes to the development of a more effective vehicle counting system for smart city applications while also paving the way for further research on vehicle detection under varying lighting and environmental conditions.

AI Powered Smart Glass for Visually Impaired Individuals

Visually impaired individuals often face difficulties navigating and interacting with their environments. This paper presents a cost-effective solution in the form of AI- powered smart glasses, which use the Arduino Uno microcontroller and ESP32 Camera to enable real-time object detection, text recognition, and obstacle avoidance. The system provides auditory and tactile feedback, making it suitable for independent navigation. The smart glasses are designed for offline functionality, utilizing edge AI to perform object and text recognition directly on the device. The paper covers the hardware design, software implementation, and evaluation of this assistive technology. Key Words: Assistive technology, Arduino Uno, ESP32 Camera, visually impaired, object detection, text-to-speech, obstacle detection, IoT.

BRTPillar: boosting real-time 3D object detection based point cloud and RGB image fusion in autonomous driving

PurposeIn autonomous driving, the inherent sparsity of point clouds often limits the performance of object detection, while existing multimodal architectures struggle to meet the real-time requirements for 3D object detection. Therefore, the main purpose of this paper is to significantly enhance the detection performance of objects, especially the recognition capability for small-sized objects and to address the issue of slow inference speed. This will improve the safety of autonomous driving systems and provide feasibility for devices with limited computing power to achieve autonomous driving.Design/methodology/approachBRTPillar first adopts an element-based method to fuse image and point cloud features. Secondly, a local-global feature interaction method based on an efficient additive attention mechanism was designed to extract multi-scale contextual information. Finally, an enhanced multi-scale feature fusion method was proposed by introducing adaptive spatial and channel interaction attention mechanisms, thereby improving the learning of fine-grained features.FindingsExtensive experiments were conducted on the KITTI dataset. The results showed that compared with the benchmark model, the accuracy of cars, pedestrians and cyclists on the 3D object box improved by 3.05, 9.01 and 22.65%, respectively; the accuracy in the bird’s-eye view has increased by 2.98, 10.77 and 21.14%, respectively. Meanwhile, the running speed of BRTPillar can reach 40.27 Hz, meeting the real-time detection needs of autonomous driving.Originality/valueThis paper proposes a boosting multimodal real-time 3D object detection method called BRTPillar, which achieves accurate location in many scenarios, especially for complex scenes with many small objects, while also achieving real-time inference speed.

Enhancing real time object detection for autonomous driving using YOLO-NAS algorithm with CLEO optimizer

Enhancing real time object detection for autonomous driving using YOLO-NAS algorithm with CLEO optimizer

Object Detection for Autonomous Vehicles Using YOLO Algorithm

The YOLO (You Only Look Once) algorithm is a real-time object recognition system that classifies objects as regression problems and predicts bounding boxes and class probabilities directly from the whole image in analysis. It is known for its speed and accuracy in real-time object detection. One of the best detection systems in autonomous vehicles is YOLO (You Only Look Once). It predicts bounding box and class probabilities from the entire image in a single evaluation using the same convolutional neural network. It works well in real-time applications such as autonomous driving due to its high-speed capability. YOLO’s one-step detection pipeline streamlines the process and reduces the computational burden. As a result, drivers and other road users are safer and have better situational awareness and decision-making skills.