Deep Learning-based Object Detection Research Articles

Probabilistic Approach for Road-Users Detection

Object detection in autonomous driving applications implies the detection and tracking of semantic objects that are commonly native to urban driving environments, as pedestrians and vehicles. One of the major challenges in state-of-the-art deep-learning based object detection are false positives which occur with overconfident scores. This is highly undesirable in autonomous driving and other critical robotic-perception domains because of safety concerns. This paper proposes an approach to alleviate the problem of overconfident predictions by introducing a novel probabilistic layer to deep object detection networks in testing. The suggested approach avoids the traditional Sigmoid or Softmax prediction layer which often produces overconfident predictions. It is demonstrated that the proposed technique reduces overconfidence in the false positives without degrading the performance on the true positives. The approach is validated on the 2D-KITTI objection detection through the YOLOV4 and SECOND (Lidar-based detector). The proposed approach enables interpretable probabilistic predictions without the requirement of re-training the network and therefore is very practical.

Detecting defects in fused deposition modeling based on improved YOLO v4

Fused deposition modeling comes with many conveniences for the manufacturing industry, but many defects tend to appear in actual production due to the problems of the FDM mechanism itself. Although some deep learning-based object detection models show excellent performance in detecting defects in the additive manufacturing process, their detection efficiency is relatively low, and they are prone to drawbacks in the face of large numbers of defects. In this paper, an improved model based on the YOLO v4 network structure is developed. We lightweight the model and modify its loss function to achieve better performance. Experimental results show that the improved model, MobileNetV2-YOLO v4, achieves a mAP of 98.96% and an FPS of 50.8 after training, which obtains higher detection accuracy and faster detection speed than the original YOLO v4 algorithm model. Through testing, this improved model can accurately identify the location and information of target defects, which has great potential for real-time detection in the additive manufacturing process.

딥러닝 기반 이미지분석 기술을 활용한 3차원 역학실험장치 개발

This research proposes to develop a device that fuses deep learning-based object detection technology and measurement algorithms to collect real-time position data of objects moving in an area. The commonly used motion analysis devices are limited to measuring two-dimensional plane motion based on images. Therefore, physical quantities such as moments of inertia and rotation are difficult to obtain. Although systems that track, measure, and analyze the motion of objects are used in sports, they are expensive and specific to the field, making it difficult to use them widely in various fields of study. Thus, this research aims to overcome these limitations by developing device that analyzes the movement of objects in various three-dimensional spaces using two vertically positioned cameras, which can be used in a variety of environments with dramatically reduced costs. The proposed device consists of two main steps: the first step is to measure the three-dimensional position of an object with a camera connected to a single-shot multibox detector (SSD) trained with 1,465 image data, and the second step is to calculate the physical quantity and analyze the motion after visualizing the data of the measured object’s position and time in three dimensions. To measure the actual coordinates of the object, the diameter of the object is converted into pixels, and the distance in centimeters per pixel is obtained by utilizing the similarity ratio with the moving object captured by the camera. With the cm/px calculated, the pixel value obtained through camera is converted to cm, the actual coordinate of the object. To verify the accuracy of the proposed data, the acceleration of gravity was measured with the trend line of each parabola. The average error range of the acceleration of gravity calculated was ±0.20 m/s², which demonstrated great accuracy.

Deep learning-based object detection for smart solid waste management system

Currently in Ethiopia, pollution and environmental damage brought on by waste increased along with industrialization, urbanization, and global population levels. Waste sorting, which is still done improperly from the household level to the final disposal site, is a prevalent issue. Real-time and accurate waste detection in image and video data is a crucial and difficult task in the intelligent waste management system. Accurately locating and classifying these wastes is challenging, particularly when there are various types of waste present. So, a single-stage YOLOv4-waste deep neural network model is proposed. In this study, a deep learning algorithm for object detection using YOLOv4 and YOLOv4-tiny is trained and evaluated. A total of 3529 waste images are divided into 7 classes, which include, cardboard, glass, metal, organic, paper, plastic, and trash. Each model uses three various inputs throughout the testing phase, including input images, videos, and webcams. Experiments with hyper-parameters on subdivision values and mosaic data augmentation were also done in the YOLOv4-tiny model. The outcome demonstrates that YOLOv4 performs better than YOLOv4-tiny for object detection specifically for waste detection. The outcome shows that YOLOv4 performs better than YOLOv4-tiny for object detection, even if YOLOv4-tiny’s scores are higher in terms of computing speed. The best results from the YOLOv4 model reach mAP 91.25%, precision 0.91, recall 0.88, F1-score 0.89, and Average IoU 81.55%, while the best YOLOv4-tiny results are mAP 82.02%, precision 0.75, recall 0.76, F1-score 0.75, and Average IoU 63.59%. This research also proves that the models with smaller subdivision values and using a mosaic have optimal performance.

A Real-Time Olive Fruit Detection for Harvesting Robot Based on YOLO Algorithms

Abstract Deep neural network models have become powerful tools of machine learning and artificial intelligence. They can approximate functions and dynamics by learning from examples. This paper reviews the state-of-art of deep learning-based object detection frameworks that are used for fruit detection in general and for olive fruit in particular. A dataset of olive fruit on the tree is built to train and evaluate deep models. The ultimate goal of this work is the capability of on-edge real-time olive fruit detection on the tree from digital videos. Recent work in deep neural networks has led to the development of a state-of-the-art object detector termed You Only Look Once version five (YOLOv5). This paper builds a dataset of 1.2 K source images of olive fruit on the tree and evaluates the latest object detection algorithms focusing on variants of YOLOv5 and YOLOR. The results of the YOLOv5 models show that the YOLOv5 new network models are able to extract rich olive features from images and detect the olive fruit with a high precision of higher than 0.75 mAP_0.5. YOLOv5s performs better for real-time olive fruit detection on the tree over other YOLOv5 variants and YOLOR.

High-Speed Tracking with Mutual Assistance of Feature Filters and Detectors.

Object detection and tracking in camera images is a fundamental technology for computer vision and is used in various applications. In particular, object tracking using high-speed cameras is expected to be applied to real-time control in robotics. Therefore, it is required to increase tracking speed and detection accuracy. Currently, however, it is difficult to achieve both of those things simultaneously. In this paper, we propose a tracking method that combines multiple methods: correlation filter-based object tracking, deep learning-based object detection, and motion detection with background subtraction. The algorithms work in parallel and assist each other's processing to improve the overall performance of the system. We named it the "Mutual Assist tracker of feature Filters and Detectors (MAFiD method)". This method aims to achieve both high-speed tracking of moving objects and high detection accuracy. Experiments were conducted to verify the detection performance and processing speed by tracking a transparent capsule moving at high speed. The results show that the tracking speed was 618 frames per second (FPS) and the accuracy was 86% for Intersection over Union (IoU). The detection latency was 3.48 ms. These experimental scores are higher than those of conventional methods, indicating that the MAFiD method achieved fast object tracking while maintaining high detection performance. This proposal will contribute to the improvement of object-tracking technology.

Improvement of Vehicle detection and classification performance with Region of Interest

This Abstract-Paper presents an innovative approach to improve the performance of vehicle detection and classification in the context of computer vision. This study proposes the use of Region of Interest (ROI) to identify and process vehicle information in traffic images. The proposed method is based on a combination of deep learning-based object detection algorithms with a more focused ROI-based approach. First, we use state-of-the-art convolution neural networks to perform object detection in traffic images, which involves a visual recognition process to identify vehicle locations. Then, using ROI, we reorder and reduce the dimensions of irrelevant data to remove background areas and increase attention to vehicle objects. This aims to reduce the computational load and speed up the detection and classification processes. We evaluate the proposed approach on a wide traffic dataset and compare the results with conventional object detection approaches without using ROI. The experimental results show that the proposed method outperforms the conventional approach in terms of vehicle detection speed and accuracy. In conclusion, this study succeeded in improving the performance of vehicle detection and classification by leveraging Region of Interest, which allows better focus on relevant regions and reduces computational complexity. These results can make an important contribution to the development of a more efficient and reliable traffic control and transportation system.

Various Approaches to Object Detection using Deep Learning

Abstract: In order to detect objects in actual scenarios, this research study examines the practical use of deep learning techniques, with a focus on convolutional neural networks (CNNs). The importance of object categorization as the fundamental building block for object detection is emphasized, and several learning strategies are investigated in order to extract features from challenging multidimensional data. In order to get accurate and effective object detection findings, the study studies various deep learning architectures, including R-CNN, Fast R-CNN, Faster R-CNN, and YOLO. In addition, this paper demonstrates the diverse uses of deep learning-based object detection, including its influence on automated vehicles, medical technology, and industrial automation, where it is changing several sectors of the economy. The analysis also foresees the future developments of deep learning models, predicting their paradigm-shifting effects across a variety of areas and their capacity to successfully address practical difficulties.

Progress in Object Detection: An In-Depth Analysis of Methods and Use Cases

Object detection, a fundamental task in computer vision, involves identifying and localizing objects within images or videos. This paper provides a comprehensive review of traditional and deep learning-based object detection techniques and their applications, challenges, and future directions. We first discuss traditional object detection methods, which rely on handcrafted features and classical machine learning algorithms. We then explore the advancements brought by deep learning, including convolutional neural networks (CNNs) and transformer-based architectures, which have significantly improved the accuracy and efficiency of object detection tasks. A thorough comparison and evaluation of different object detection techniques are presented, considering performance metrics, speed, and robustness to object size, orientation, and occlusion variations. We also examine the diverse applications of object detection across various domains, such as robotics, autonomous vehicles, surveillance, medical imaging, and augmented reality. We outline open challenges and future research directions, emphasizing the need to combine object detection with other tasks, develop few-shot and zero-shot learning approaches, and address issues related to fairness, accountability, and transparency. This paper aims to comprehensively review the most prominent object detection techniques, their evolution, and their applications in diverse domains. We discussed traditional methods and recent deep learning-based approaches, emphasizing their strengths and limitations.

Deep learning based object detection from multi-modal sensors: an overview

Deep learning based object detection from multi-modal sensors: an overview

Don’t get lost in the crowd: Graph convolutional network for online animal tracking in dense groups

Tracking the behaviour of animals in group-housed situations is a critical area of study for precision livestock farming, but it poses challenges in diverse and crowded environments. Existing methods often struggle with false positives and false negatives due to the complexities of these scenarios. In this study, we propose a robust computer vision algorithm for long-term (>10mins) animal tracking, with a primary focus on group-housed pigs. Our method addresses the limitations of current approaches by effectively handling errors from the detector in challenging environments. Our approach integrates Graph Convolutional Networks.(GCNs) with deep learning-based object detection techniques. We represent the data as a graph structure, with nodes corresponding to multiple animal detections over several frames. Edges connect thedetections that do not appear in the same frame. The model's objective is to perform edge classification, where each edge is associated with a scalar representing the probability of being the same object. To enhance the robustness of edge classification, we combine the classifier withan ensemble predictor, enabling accurate tracking of animal identitiesfor extended periods. Notably, our graph convolutional model achievesaccurate re identification of non-consecutive animal detections inreal-time without the need for trajectory prediction or ID association. We evaluate our method by comparing its performance against othertracking methods, including DeepSORT—a state-of-the-art animal trackingmethod that relies on a Kalman filter, a deep appearance descriptor, and the Hungarian algorithm for trajectory prediction and ID association.Our comparison demonstrates the advantages of our method, as it (1)improves the IDF1 score, indicating enhanced ID association accuracy, by 1.72%; (2) mitigates 93% of errors resulting from ID-switch anddeviation; (3) extends the tracking duration of each animal byapproximately 66% in challenging conditions; and (4) achieves aprocessing speed that is 1000 times faster than DeepSORT when operatingat 22fps with the detector. The potential applications of our method extend to various aspects of livestock management. By accurately tracking individual animals, we can monitor behaviours such as feeding, drinking, and aggressive interactions. This information can lead to improvements in animal welfare, resource allocation, and breeding strategies. Follow-ups of this research are available at: https://gitlab.kuleuven.be/m3-biores/public/m3pig.

Scalable and effective artificial intelligence for multivariate radar environment

Ground based radar is a key system for aerial defence of any country. Current generations of radar use Moving Target Indicator (MTI) filter, Constant False Alarm Rate (CFAR) and Peak Detection (PD) algorithms for detection of target and rejection of non-target (static clutter, dynamic clutter, interference and noise) objects. These conventional Radar Signal Processing (RSP) techniques find their limitation when either targets are not detected or false alarms are generated due to lack of built-in intelligence. In order to eliminate both the issues, a realistic radar environment has been analyzed using Artificial Intelligence (AI). Special emphasis has been made on invalid detections caused by weather clutter, which was never worked upon previously using AI. A novel deep learning based object detection approach has been devised keeping radar environment, high accuracy requirement and real-time constraint of radar in view. The proposed approach has been validated for its efficacy through standalone testing as well as Poof of Concept (PoC) implementation on a real radar. Integration of AI model in radar platform also involves a novel implementation scheme, whereby results of AI model are augmented by conventional RSP. Inference time of AI models is a bottleneck for their use in radar systems, in this context, new strategies have also been analyzed to achieve a real-time performance for radar application. As the objective of this research is to devise scalable and effective solutions for upcoming generations of radar; therefore, the work is of practical value for realistic deployments.

Single-stage object detector with local binary pattern for remote sensing images

ABSTRACT The objects in remote sensing images usually have complex background and appear anywhere in any direction, which poses challenges for deep learning-based object detection. In order to solve the above problems, we embed the local binary pattern (LBP) into deep learning-based object detection to facilitate performance improvements. Specifically, we implement thep LBP via depthwise separable convolution without any learnable parameters, which makes LBP highlight the object-related regions of feature map instead of original images. The convolutional implementation of LBP makes it easy to embed it in deep learning-based object detection, so that deep learning-based object detection can possess the ability of LBP to extract the texture information from complex background without having to learn. The gradients of the branch where LBP convolution is located are blocked and cannot be backpropagated according to proposed LBP convolution, so we propose an effective residual block, called LBPRes (Residual block with Local Binary Pattern) for normal training. To deal with the problem of object direction changes, we introduce the rotation invariance into the LBPRes, denoted as RILBPRes (LBPRes with Rotation Invariance), which makes deep learning-based object detection have the local rotation invariance to cope with the object direction changes. Finally, we construct a single-stage object detection network S2LBPNet (Single Stage object detection Network with Local Binary Pattern) with a local binary pattern and conduct related experiments on the DIOR and HRRSD datasets. According to the experimental results, with a small number of parameters added, S2LBPNet outperforms the YOLOv5s baseline by 2.2% and 5.1% on DIOR and HRRSD datasets, respectively, which prove the superiority of the proposed S2LBPNet.

Investigating Failure Patterns in Machine Learning-based Object Detection Tasks in Software Development Courses

<p>Object detection, one of the popular tasks in computer vision, is to find all objects of interest in an image and determine their category and location. When people use deep learning frameworks to implement object detection networks, defects are often caused by human-introduced faults. These defects may cause different types of failures. Exploring frequent failure patterns in object detection programs can help developers detect and fix defects more effectively and efficiently. Therefore, we conducted an empirical study on failure patterns in deep learning-based object detection programs submitted in university software development courses. By exploring 101 submissions of a Yolov4 object detection task completed by 104 students, we found the most frequent 13 failure patterns in these submissions and six types of root causes of these failures. To help students and entry-level software engineers avoid possible faults in object detection programs, 13 concrete suggestions that belong to six classes are given in this paper. These results can reveal some basic laws of failures and mistakes in the development of deep learning-based object detection programs and provide guidances to assist students and entry-level developers in improving their skills in developing object detection programs. </p> <p> </p>

Identification and Tracking of Groups of People using Object Detection and Object Tracking

Object detection is one of the most important areas in the fields of Data Science and Computer Vision. In this paper, we present a novel approach to identifying and tracking groups of people, couples, and individuals in videos by using deep learning-based object detection and object tracking techniques along with a proposed grouping algorithm. In this approach, transfer learning is applied on YOLO v3 model for the detection of people in video frames, and Deep SORT is applied for tracking each detected person throughout the video. Results obtained from person detection and person tracking were used by the proposed grouping algorithm to identify and track groups, couples, and individuals who are appearing in input videos. Our proposed grouping algorithm is based on the proximity between each individual and the time duration that proximity is maintained for. It also considers how to identify and track groups, when people are moving within the groups. This approach was evaluated using CCTV videos captured from the restaurant domain and it was able to perform the group detection and tracking tasks successfully with a precision of 0.7083, recall of 0.7906 and F1 score of 0.7471.

State-of-the-Art Deep Learning Methods for Objects Detection in Remote Sensing Satellite Images.

Introduction: Object detection in remotely sensed satellite images is critical to socio-economic, bio-physical, and environmental monitoring, necessary for the prevention of natural disasters such as flooding and fires, socio-economic service delivery, and general urban and rural planning and management. Whereas deep learning approaches have recently gained popularity in remotely sensed image analysis, they have been unable to efficiently detect image objects due to complex landscape heterogeneity, high inter-class similarity and intra-class diversity, and difficulty in acquiring suitable training data that represents the complexities, among others. Methods: To address these challenges, this study employed multi-object detection deep learning algorithms with a transfer learning approach on remotely sensed satellite imagery captured on a heterogeneous landscape. In the study, a new dataset of diverse features with five object classes collected from Google Earth Engine in various locations in southern KwaZulu-Natal province in South Africa was used to evaluate the models. The dataset images were characterized with objects that have varying sizes and resolutions. Five (5) object detection methods based on R-CNN and YOLO architectures were investigated via experiments on our newly created dataset. Conclusions: This paper provides a comprehensive performance evaluation and analysis of the recent deep learning-based object detection methods for detecting objects in high-resolution remote sensing satellite images. The models were also evaluated on two publicly available datasets: Visdron and PASCAL VOC2007. Results showed that the highest detection accuracy of the vegetation and swimming pool instances was more than 90%, and the fastest detection speed 0.2 ms was observed in YOLOv8.

An early warning model for starfish disaster based on multi-sensor fusion

Starfish have a wide range of feeding habits, including starfish, sea urchins, sea cucumbers, corals, abalones, scallops, and many other marine organisms with economic or ecological value. The starfish outbreak in coastal areas will lead to severe economic losses in aquaculture and damage the ecological environment. However, the current monitoring methods are still artificial, time-consuming, and laborious. This study used an underwater observation platform with multiple sensors to observe the starfish outbreak in Weihai, Shandong Province. The platform could collect the temperature, salinity, depth, dissolved oxygen, conductivity, other water quality data, and underwater video data. Based on these data, the paper proposed an early warning model for starfish prevalence (EWSP) based on multi-sensor fusion. A deep learning-based object detection method extracts time-series information on the number of starfish from underwater video data. For the extracted starfish quantity information, the model uses the k-means clustering algorithm to divide the starfish prevalence level into four levels: no prevalence, mild prevalence, medium prevalence, and high prevalence. Correlation analysis concluded that the water quality factors most closely related to the starfish prevalence level are temperature and salinity. Therefore, the selected water quality factor and the number of historical starfish are inputted. The future starfish prevalence level of the starfish outbreak is used as an output to train the BP (back propagation) neural network to build EWSP based on multi-sensor fusion. Experiments show that the accuracy rate of this model is 97.26%, whose precision meets the needs of early warning for starfish outbreaks and has specific application feasibility.

LKC-Net: large kernel convolution object detection network

Deep learning-based object detection methods have achieved great performance improvement. However, since small kernel convolution has been widely used, the semantic feature is difficult to obtain due to the small receptive fields, and the key information cannot be highlighted, resulting in a series of problems such as wrong detection, missing detection, and repeated detection. To overcome these problems, we propose a large kernel convolution object detection network based on feature capture enhancement and vast receptive field attention, called LKC-Net. Firstly, a feature capture enhancement block based on large kernel convolution is proposed to improve the semantic feature capturing ability, and depth convolution is used to reduce the number of parameters. Then, the vast receptive filed attention mechanism is constructed to enhance channel direction information extraction ability, and it is more compatible with the proposed backbone than other existing attention mechanisms. Finally, the loss function is improved by introducing the SIoU, which can overcome the angle mismatch problem between the ground truth and prediction box. Experiments are conducted on Pascal VOC and MS COCO datasets for demonstrating the performance of LKC-Net.

Multi-Scale Feature Selective Matching Network for Object Detection

Numerous deep learning-based object detection methods have achieved excellent performance. However, the performance on small-size object detection and positive and negative sample imbalance problems is not satisfactory. We propose a multi-scale feature selective matching network (MFSMNet) to improve the performance of small-size object detection and alleviate the positive and negative sample imbalance problems. First, we construct a multi-scale semantic enhancement module (MSEM) to compensate for the information loss of small-sized targets during down-sampling by obtaining richer semantic information from features at multiple scales. Then, we design the anchor selective matching (ASM) strategy to alleviate the training dominated by negative samples caused by the imbalance of positive and negative samples, which converts the offset values of the localization branch output in the detection head into localization scores and reduces negative samples by discarding low-quality anchors. Finally, a series of quantitative and qualitative experiments on the Microsoft COCO 2017 and PASCAL VOC 2007 + 2012 datasets show that our method is competitive compared to nine other representative methods. MFSMNet runs on a GeForce RTX 3090.

Deep Learning-Based Image Classification for Major Mosquito Species Inhabiting Korea.

Mosquitoes are one of the deadliest insects, causing harm to humans worldwide. Preemptive prevention and forecasting are important to prevent mosquito-borne diseases. However, current mosquito identification is mostly conducted manually, which consumes time, wastes labor, and causes human error. In this study, we developed an automatic image analysis method to identify mosquito species using a deep learning-based object detection technique. Color and fluorescence images of live mosquitoes were acquired using a mosquito capture device and were used to develop a deep learning-based object detection model. Among the deep learning-based object identification models, the combination of a swine transformer and a faster region-convolutional neural network model demonstrated the best performance, with a 91.7% F1-score. This indicates that the proposed automatic identification method can be rapidly applied for efficient analysis of species and populations of vector-borne mosquitoes with reduced labor in the field.