Object Detection Problem Research Articles

Automated Design of Salient Object Detection Algorithms with Brain Programming

Despite recent improvements in computer vision, artificial visual systems’ design is still daunting since an explanation of visual computing algorithms remains elusive. Salient object detection is one problem that is still open due to the difficulty of understanding the brain’s inner workings. Progress in this research area follows the traditional path of hand-made designs using neuroscience knowledge or, more recently, deep learning, a particular branch of machine learning. Recently, a different approach based on genetic programming appeared to enhance handcrafted techniques following two different strategies. The first method follows the idea of combining previous hand-made methods through genetic programming and fuzzy logic. The second approach improves the inner computational structures of basic hand-made models through artificial evolution. This research proposes expanding the artificial dorsal stream using a recent proposal based on symbolic learning to solve salient object detection problems following the second technique. This approach applies the fusion of visual saliency and image segmentation algorithms as a template. The proposed methodology discovers several critical structures in the template through artificial evolution. We present results on a benchmark designed by experts with outstanding results in an extensive comparison with the state of the art, including classical methods and deep learning approaches to highlight the importance of symbolic learning in visual saliency.

Stepwise Domain Adaptation (SDA) for Object Detection in Autonomous Vehicles Using an Adaptive CenterNet

In recent years, deep learning technologies for object detection have made great progress and have powered the emergence of state-of-the-art models to address object detection problems. Since the domain shift can make detectors unstable or even crash, the detection of cross-domain becomes very important for the design of object detectors. However, traditional deep learning technologies for object detection always rely on a large amount of reliable ground-truth labelling that is laborious, costly, and time-consuming. Although an advanced approach CycleGAN has been proposed for cross-domain object detection tasks, the ability of CycleGAN to reduce the divergence across domains at the feature level is limited. In this paper, a stepwise domain adaptation (SDA) detection method is proposed to further improve the performance of CycleGAN by minimizing the divergence in cross-domain object detection tasks. Specifically, the domain shift is addressed in two steps. In the first step, to bridge the domain gap, an unpaired image-to-image translator is trained to construct a fake target domain by translating the source images to the similar ones in the target domain. In the second step, to further minimize divergence across domains, an adaptive CenterNet is designed to align distributions at the feature level in an adversarial learning manner. Our proposed method is evaluated in domain shift scenarios based on the driving datasets including Cityscapes, Foggy Cityscapes, SIM10k, and BDD100K. The results show that our method is superior to the state-of-the-art methods and is effective for object detection in domain shift scenarios.

DESAIN MODEL KLASIFIKASI SAMPAH ORGANIK MENJADI BAHAN BAKU BRIKET BIOMASSA MENGGUNAKAN METODE DEEP LEARNING

The development of information technology in Makassar has implemented the smart city concept. The mission of urban development is to build the economy of urban communities that have superior capacity, competitiveness, productivity, creativity, innovation, efficiency and use information technology to utilize the waste generated by the community and companies. Cities have problems stemming from development by producing waste in the form of organic and inorganic waste. To empower the community, city governance is needed with a design concept that continues towards green technology. The aim of the research is to design a model that can sort and identify waste based on various types of organic matter that can be used as raw material for making biomass briquettes as a renewable alternative energy source. Research uses deep learning methods as part of machine learning in solving object detection problems in digital image classification in computer vision. The research resulted in a classification model design based on object detection by classifying 5 types of organic waste.

Workshop Safety Helmet Wearing Detection Model Based on SCM-YOLO.

In order to overcome the problems of object detection in complex scenes based on the YOLOv4-tiny algorithm, such as insufficient feature extraction, low accuracy, and low recall rate, an improved YOLOv4-tiny safety helmet-wearing detection algorithm SCM-YOLO is proposed. Firstly, the Spatial Pyramid Pooling (SPP) structure is added after the backbone network of the YOLOv4-tiny model to improve its adaptability of different scale features and increase its effective features extraction capability. Secondly, Convolutional Block Attention Module (CBAM), Mish activation function, K-Means++ clustering algorithm, label smoothing, and Mosaic data enhancement are introduced to improve the detection accuracy of small objects while ensuring the detection speed. After a large number of experiments, the proposed SCM-YOLO algorithm achieves a mAP of 93.19%, which is 4.76% higher than the YOLOv4-tiny algorithm. Its inference speed reaches 22.9FPS (GeForce GTX 1050Ti), which meets the needs of the real-time and accurate detection of safety helmets in complex scenes.

Long-Distance Vehicle Dynamic Detection and Positioning Based on Gm-APD Lidar and LIDAR-YOLO

Aiming at the problems of long-distance night dynamic object detection in military operations, such as difficult imaging of traditional equipment, a small number of object pixels, a large difference in object size at different distances, and poor real-time performance. A long-distance vehicle detection method based on the LIDAR-YOLO network for Gm-APD (Geiger mode avalanche photodiode) lidar image is proposed. This method is improved based on YOLOv5s, which can detect long-distance vehicles and accurately locate motion trajectories in real-time. Firstly, the intensity image I of the lidar is preprocessed and combined with the distance image R, the BC_IIR image is obtained, and the vehicle detection data set is established. Then, according to the obvious vehicle features in the BC_IIR image, the low-complexity backbone network LIDARNet is redesigned, and several attention mechanism modules are added. The network feature extraction ability is enhanced by comparing and selecting the CA attention mechanism module. In the network prediction layer, context-aware feature map fusion is added to improve the detection precision of vehicles at different distances. Next, a GIOU loss is introduced to improve the detection precision of occluded vehicles. Finally, the method can depict the trajectory of the vehicle in 3D space based on distance information. After testing, the network achieves a detection speed of 62.67 frames per second (FPS), recall rate (R) = 98.19%, precision rate (P) = 95.53%, and F1 is 0.9675. The results show that the algorithm has better performance than other methods.

Improved YOLOv5 Based on Hybrid Domain Attention for Small Object Detection in Optical Remote Sensing Images

The object detection technology of optical remote sensing images has been widely applied in military investigation, traffic planning, and environmental monitoring, among others. In this paper, a method is proposed for solving the problem of small object detection in optical remote sensing images. In the proposed method, the hybrid domain attention units (HDAUs) of channel and spatial attention mechanisms are combined and employed to improve the feature extraction capability and suppress background noise. In addition, we designed a multiscale dynamic weighted feature fusion network (MDW-Net) to improve adaptive optimization and deep fusion of shallow and deep feature layers. The model is trained and tested on the DIOR dataset, and some ablation and comparative experiments are carried out. The experimental results show that the mAP of the proposed model surpasses that of YOLOv5 by a large margin of +2.3 and has obvious advantages regarding the detection performance for small object categories, such as airplane, ship, and vehicle, which support its application for small target detection in optical remote sensing images.

Improved YOLOv5-S object detection method for optical remote sensing images based on contextual transformer

To address the problems of error and omission detection in remote sensing image detection caused by the diverse scale changes of remote sensing object scales and the abundant proportion of small-scale objects, as well as the global and dense distribution of remote sensing objects, a remote sensing image detection improvement method based on YOLOv5-S is proposed. First, according to the characteristics of remote sensing objects, the data enhancement strategy is adopted to expand the dataset samples for the characteristics of remote sensing objects to improve the generalization ability of the model. Second, the contextual transformer module is introduced to the backbone feature extraction network and the feature fusion network to ensure the local feature extraction capability while improving the global information acquisition capability of the model, making full use of the input contextual information and guiding the dynamic attention matrix learning to improve the visual representation ability. Third, based on the original model, a shallow detection scale is added, and then a multiscale complex fusion structure is adopted. Meanwhile, the K-means++ algorithm replaces the original K-means algorithm and then clusters 12 anchor box sizes. Fourth, the efficient intersection over union loss is used to improve the accuracy of the remote sensing object recognition prediction. In the experiment on the on two optical remote sensing image datasets, a comparison with several object detection algorithms based on convolutional neural network is made, the results show that the mAP@0.5 tested on the remote sensing datasets is higher than the original YOLOv5-S. Compared with other models, the detection efficiency is higher, and the problems of small-scale object detection in remote sensing image have been significantly improved.

Attention-Based Multi-Level Feature Fusion for Object Detection in Remote Sensing Images

We study the problem of object detection in remote sensing images. As a simple but effective feature extractor, Feature Pyramid Network (FPN) has been widely used in several generic vision tasks. However, it still faces some challenges when used for remote sensing object detection, as the objects in remote sensing images usually exhibit variable shapes, orientations, and sizes. To this end, we propose a dedicated object detector based on the FPN architecture to achieve accurate object detection in remote sensing images. Specifically, considering the variable shapes and orientations of remote sensing objects, we first replace the original lateral connections of FPN with Deformable Convolution Lateral Connection Modules (DCLCMs), each of which includes a 3×3 deformable convolution to generate feature maps with deformable receptive fields. Additionally, we further introduce several Attention-based Multi-Level Feature Fusion Modules (A-MLFFMs) to integrate the multi-level outputs of FPN adaptively, further enabling multi-scale object detection. Extensive experimental results on the DIOR dataset demonstrated the state-of-the-art performance achieved by the proposed method, with the highest mean Average Precision (mAP) of 73.6%.

Research on Intelligent Video Detection of Small Targets Based on Deep Learning Intelligent Algorithm.

Compared with the traditional object detection algorithm, the object detection algorithm based on deep learning has stronger robustness to complex scenarios, which is the hot direction of current research. According to the process characteristics of the object detection algorithm based on deep learning, it is divided into two-stage object detection algorithm and single-stage object detection algorithm, focusing on the problems solved by some classical algorithms and their advantages and disadvantages. In view of the problem of object detection, especially small object detection, the commonly used data sets and performance evaluation indicators are summarized; the characteristics, advantages, and detection difficulties of various common data sets are compared; the challenges faced by commonly used object detection methods and small object detection are systematically summarized; the latest work of small object detection methods based on deep learning is sorted out; and the small object detection methods based on multiscale and small object detection methods based on super-resolution are introduced. At the same time, the lightweight strategy for target detection methods and the performance of some lightweight models are introduced; the characteristics, advantages, and limitations of various methods are summarized; and the future development direction of small object detection methods based on deep learning is prospected.

Static-Dynamic Co-teaching for Class-Incremental 3D Object Detection

Deep learning-based approaches have shown remarkable performance in the 3D object detection task. However, they suffer from a catastrophic performance drop on the originally trained classes when incrementally learning new classes without revisiting the old data. This "catastrophic forgetting" phenomenon impedes the deployment of 3D object detection approaches in real-world scenarios, where continuous learning systems are needed. In this paper, we study the unexplored yet important class-incremental 3D object detection problem and present the first solution - SDCoT, a novel static-dynamic co-teaching method. Our SDCoT alleviates the catastrophic forgetting of old classes via a static teacher, which provides pseudo annotations for old classes in the new samples and regularizes the current model by extracting previous knowledge with a distillation loss. At the same time, SDCoT consistently learns the underlying knowledge from new data via a dynamic teacher. We conduct extensive experiments on two benchmark datasets and demonstrate the superior performance of our SDCoT over baseline approaches in several incremental learning scenarios. Our code is available at https://github.com/Na-Z/SDCoT.

DLAFS CASCADE R-CNN: AN OBJECT DETECTOR BASED ON DYNAMIC LABEL ASSIGNMENT

Object detection methods based on Deep Learning are the revolution of the Computer Vision field in general and object detection problems in particular. In detail, they are methods that belonged to the R - CNN family: Faster R - CNN and Cascade R - CNN. The characteristic of them is the Region Proposal Network, which is utilized for generating proposal regions that may include objects or not, then the proposals will be classified by the IoU threshold. In this study, we apply dynamic training, which adjusts this IoU threshold depending on the statistic of proposal regions on the Faster R - CNN and Cascade R - CNN, training on the SeaShips and DODV dataset. Cascade R - CNN with dynamic training achieve higher results compared to normal on both two datasets (higher 0.2% and 5.7% on the SeaShips and DODV dataset, respectively). In the DODV dataset, Faster R - CNN with dynamic training also perform higher results compared to its normal version, 4.4% higher.

DPSSD: Dual-Path Single-Shot Detector.

Object detection is one of the most important and challenging branches of computer vision. It has been widely used in people’s lives, such as for surveillance security and autonomous driving. We propose a novel dual-path multi-scale object detection paradigm in order to extract more abundant feature information for the object detection task and optimize the multi-scale object detection problem, and based on this, we design a single-stage general object detection algorithm called Dual-Path Single-Shot Detector (DPSSD). The dual path ensures that shallow features, i.e., residual path and concatenation path, can be more easily utilized to improve detection accuracy. Our improved dual-path network is more adaptable to multi-scale object detection tasks, and we combine it with the feature fusion module to generate a multi-scale feature learning paradigm called the “Dual-Path Feature Pyramid”. We trained the models on PASCAL VOC datasets and COCO datasets with 320 pixels and 512 pixels input, respectively, and performed inference experiments to validate the structures in the neural network. The experimental results show that our algorithm has an advantage over anchor-based single-stage object detection algorithms and achieves an advanced level in average accuracy. Researchers can replicate the reported results of this paper.

AlexNet architecture based convolutional neural network for toxic comments classification

Today online networking has become an indispensable part of life for people all over the world. It is difficult for users to reduce their internet/online communications, as the flow of information increases everyday. While the free flow of information benefits online communications, the high toxicity of online communication is a drawback. Toxic texts are described as disrespectful or insulting messages that make the recipient feel uncomfortable. Deep Learning based Convolutional Neural Networks (CNN) have given exceptional outcomes in Computer Vision Domain, and AlexNet has proven to be the leading architecture in image classification and object detection problems. This article presents a 3-tier CNN architecture that is inspired by the AlexNet model to classify the toxic comments on the Wikipedia forum available in the Google Jigsaw dataset. Fast text-crawl-300d-2 m is used to formulate the pre-trained word embeddings matrix. The Exponential Linear Unit (ELU) activation function is applied in the Convolutional blocks for faster convergence. Dropout is used sufficiently along with different layers of the network to prevent overfitting. From the simulation and subsequent comparative analysis, it is found that the proposed model achieved a decent average accuracy of 98.505% and an average F1 score of 0.79. ROC-AUC score is used as an evaluation parameter. The value of ROC-AUC for the proposed model is approximately 0.9854, which shows that the said model differentiates between the comment classes more accurately.

Learning to Detect Open Carry and Concealed Object With 77 GHz Radar

Detecting harmful carried objects plays a key role in intelligent surveillance systems and has widespread applications, for example, in airport security. In this paper, we focus on the relatively unexplored area of using low-cost 77GHz mmWave radar for the carried objects detection problem. The proposed system is capable of real-time detecting three classes of objects - laptop, phone, and knife - under open carry and concealed cases where objects are hidden with clothes or bags. This capability is achieved by the initial signal processing for localization and generating range-azimuth-elevation image cubes, followed by a deep learning-based prediction network and a multi-shot post-processing module for detecting objects. Extensive experiments for validating the system performance on detecting open carry and concealed objects have been presented with a self-built radar-camera testbed and collected dataset. Additionally, the influence of different input formats, factors, and parameters on system performance is analyzed, providing an intuitive understanding of the system. This system would be the very first baseline for other future works aiming to detect carried objects using 77GHz radar.

Data-Centric Learning Method for Synthetic Data Augmentation and Object Detection

This paper proposes a deep learning framework for decreasing large-scale domain shift problems in object detection using domain adaptation techniques. We have approached data-centric domain adaptation with Image-to-Image translation models for this problem. It is one of the methodologies that changes source data to target domain's style by reducing domain shift. However, the method cannot be applied directly to the domain adaptation task because the existing Image-to-Image model focuses on style translation. We solved this problem using the data-centric approach simply by reordering the training sequence of the domain adaptation model. We defined the features to be content and style. We hypothesized that object-specific information in images was more closely tied to the content than the style and thus experimented with methods to preserve content information before style was learned. We trained the model separately only by altering the training data. Our experiments confirmed that the proposed method improves the performance of the domain adaptation model and increases the effectiveness of using the generated synthetic data for training object detection models. We compared our approach with the existing single-stage method where content and style were trained simultaneously. We argue that our proposed method is more practical for training object detection models than others. The emphasis in this study is to preserve image content while changing the style of the image. In the future, we plan to conduct additional experiments to apply synthetic data generation technology to various other application areas like indoor scenes and bin picking.

A System for a Real-Time Electronic Component Detection and Classification on a Conveyor Belt

The presented research addresses the real-time object detection problem with small and moving objects, specifically the surface-mount component on a conveyor. Detecting and counting small moving objects on the assembly line is a challenge. In order to meet the requirements of real-time applications, state-of-the-art electronic component detection and classification algorithms are implemented into powerful hardware systems. This work proposes a low-cost system with an embedded microcomputer to detect surface-mount components on a conveyor belt in real time. The system detects moving, packed, and unpacked surface-mount components. The system’s performance was experimentally investigated by implementing several object-detection algorithms. The system’s performance with different algorithm implementations was compared using mean average precision and inference time. The results of four different surface-mount components showed average precision scores of 97.3% and 97.7% for capacitor and resistor detection. The findings suggest that the system with the implemented YOLOv4-tiny algorithm on the Jetson Nano 4 GB microcomputer achieves a mean average precision score of 88.03% with an inference time of 56.4 ms and 87.98% mean average precision with 11.2 ms inference time on the Tesla P100 16 GB platform.

WINDMILLS DETECTION USING DEEP LEARNING ON SENTINEL SATELLITE IMAGES

Abstract. Automatic detection of objects from Earth Observation images is still a challenge for researchers. This paper aims at extracting automatically windmills on mid-resolution images (10-meter resolution), based on Sentinel satellite products. Sentinel-2 optical images are obvious candidates for our study. At 10-meter resolution, a windmill is represented with only a few pixels. We also start to consider Synthetic Aperture Radar (SAR) Sentinel-1 images but no particular windmill radar response on GRD (Ground Range Multi-look Detected) products seemed to be relevant.Considering the maturity of deep learning techniques for object detection in computer vision, we explore the use of deep neural networks for windmill detection on remote sensing images. For that purpose, we had to create the training data sets but we took advantage of the availability of many Sentinel images and of the use of automated labelling as the objects are georeferenced. The proposed approach relies on the U-Net framework, reformulating our problem of object detection in terms of semantic segmentation. We trained several neural networks on different data sets emanating from different countries. That enabled us to measure the performance of detection within a country but also across two countries (training on a country and predicting on another country). The results show the ability of detection of such small objects with respect to the resolution and we obtain various levels of performances depending on the trained and test data sets.

Residual objectness for imbalance reduction

As most object detectors rely on dense candidate samples to cover objects, they have always suffered from the extreme imbalance between very few foreground samples and numerous background samples during training, i.e., the foreground-background imbalance. Although several resampling and reweighting schemes (e.g., OHEM, Focal Loss, GHM) have been proposed to alleviate the imbalance, they are usually heuristic with multiple hyper-parameters, which is difficult to generalize on different object detectors and datasets. In this paper, we propose a novel Residual Objectness (ResObj) mechanism that adaptively learns how to address the foreground-background imbalance problem in object detection. Specifically, we first formulate the imbalance problems on all object classes as an imbalance problem on an “objectness” class. Then, we design multiple cascaded objectness estimators with residual connections for that objectness class to progressively distinguish the foreground samples from background samples. With our residual objectness mechanism, object detectors can learn how to address the foreground-background problem in an end-to-end way, rather than rely on hand-crafted resampling or reweighting schemes. Extensive experiments on the COCO benchmark demonstrate the effectiveness and compatibility of our method for various object detectors: the RetinaNet-ResObj, YOLOv3-ResObj and FasterRCNN-ResObj achieve relative 3%∼4% Average Precision (AP) improvements compared with their vanilla models, respectively.

MeterSSD: Scale and rotate invariance lightweight detector

Since traditional CNNs do not have explicit rotation and scale invariance, it is difficult to adapt to the scale changes of objects in real scenes when dealing with object detection tasks. There have been many researches on rotating object detection, which have achieved outstanding performance. But most of these works focus on anchor design, such as different multi-scale anchors, rotating BBox and rotating ROI Pooling. Although these methods have solved some practical problems in reality, the methods increase the cost of box generation significantly. We are committed to designing a rotation and scale invariance detector. A lightweight and efficient detector MeterSSD we proposed is based on spatial attention, channel attention and linear transformation. We not only decrease the rotation transformation cost but also ensure the detector to be able to learn multi-scale features. We improve MobileNetV2 and use it as the backbone for SSDLite. Adaptive calibration of feature channels concentrates more on specific features. In order to compress MeterSSD, we replace partial convolution operations with linear transformation operations. We apply spatial attention to the detection head to obtain more orientation and scale invariance of feature map. MeterSSD significantly improve the deformable detection accuracy of SSD without speed sacrifice. It can be applied to general object detection problems, and reaches 65.53% mAP on Pascal VOC2012 and 22.74% mAP on COCO2017. Furthermore, we research the field of automatic meter reading and propose a large-scale annotated datasets GPT-AMR, which contains 78,091 water meter images from different manufacturers and different types. Our MeterSSD achieves 90.00% mAP at 52 fps on GPT-AMR datasets. The overall accuracy of digit recognition is 97.16%. Experiments show that MeterSSD enhance accuracy slightly compared to the current SOTA detector dose, but improves the detection accuracy of deformable objects prominently, and can adapt to any scene without modification.

The fishmeal adulteration identification based on microscopic image and deep learning

The fishmeal adulteration identification based on microscopic image and deep learning was studied in this paper, which included the qualitative identification and component recognition of adulterated fishmeal. Mobilenetv2 was chosen as the qualitative identification model to distinguish between fishmeal samples and adulterated samples. And the accuracy, precision, recall, and F1-measure of the test set were adopted as the evaluation indicators. The YOLOv3-Mobilenetv2 was selected as the component recognition model to distinguish components in adulterated samples, in which the feature pyramid structure and multi-scale feature fusion strategy were applied to solve the multi-scale problem in object detection. And the mean average precision (mAP) was adopted as the precision evaluation index and the frame per second (FPS) as the speed evaluation index of the component recognition model. At the same time, the effects of image enhancement preprocessing and resolution on the identification of fishmeal adulteration were also considered in this paper. It was shown that Mobilenetv2 was more suitable for qualitative identification when using the local constrained mask transformed microscopic images with a resolution of 224 × 224 pixels. The average accuracy, precision, recall, and F1-measure of the model were 93.02%, 93.11%, 93.25%, and 92.69% respectively. The improved YOLOv3-Mobilenetv2 was more suitable for component recognition with a resolution of 608 × 608 pixels. The mAP reached 78.49%, and the FPS was 45.97 f·s−1. The model of fishmeal adulteration identification adopted in this paper had high accuracy, which could quickly identify adulterated fishmeal and recognize components, and provide a new method and reference for improving the objectivity of fishmeal adulteration identification.