Accurate Object Detector Research Articles

DAN-YOLO: A Lightweight and Accurate Object Detector Using Dilated Aggregation Network for Autonomous Driving

Object detection is becoming increasingly critical in autonomous driving. However, the accuracy and effectiveness of object detectors are often constrained by the obscuration of object features and details in adverse weather conditions. Therefore, this paper presented the DAN-YOLO vehicle object detector specifically designed for driving conditions in adverse weather. Building on the YOLOv7-Tiny network, SPP was replaced with SPPF, resulting in the SPPFCSPC structure, which enhances processing speed. The concept of Hybrid Dilated Convolution (HDC) was also introduced to improve the SPPFCSPC and ELAN-T structures, expanding the network’s receptive field (RF) while maintaining a lightweight design. Furthermore, an efficient multi-scale attention (EMA) mechanism was introduced to enhance the effectiveness of feature fusion. Finally, the Wise-IoUv1 loss function was employed as a replacement for CIoU to enhance the localization accuracy of the bounding box (bbox) and the convergence speed of the model. With an input size of 640 × 640, the DAN-YOLO algorithm proposed in this study achieved an increase in mAP0.5 values of 3.4% and 6.3% compared to the YOLOv7-Tiny algorithm in the BDD100K and DAWN benchmark tests, respectively, while achieving real-time detection (142.86 FPS). When compared with other state-of-the-art detectors, it reports better trade-off in terms of detection accuracy and speed under adverse driving conditions, indicating the suitability for autonomous driving applications.

LHFNet: A Fast and Accurate Object Detector Based on CenterNet

This paper proposes LHFNet, an improved object detection method based on CenterNet, which achieves a better speed-accuracy trade-off. The main contributions of this paper are as follows: We design a lightweight Hourglass network to reduce the number of parameters and computations of CenterNet, thus improving the detection speed. We introduce an intermediate feature map fusion module to enhance the feature extraction capability of Hourglass, which compensates for the possible performance degradation caused by the network simplification. We apply an attention module to the final heat map to improve the image feature representation. We design and optimize a series of loss functions to train LHFNet effectively. Through the research in this paper, we demonstrate that the feature map fusion and the attention module greatly improve the performance of LHFNet. We evaluate our method on the COCO dataset and achieve a high accuracy at a fast speed (Tesla V100), which outperforms other object detection methods in terms of both accuracy and efficiency.

Point-Teaching: Weakly Semi-supervised Object Detection with Point Annotations

Point annotations are considerably more time-efficient than bounding box annotations. However, how to use cheap point annotations to boost the performance of semi-supervised object detection is still an open question. In this work, we present Point-Teaching, a weakly- and semi-supervised object detection framework to fully utilize the point annotations. Specifically, we propose a Hungarian-based point-matching method to generate pseudo labels for point-annotated images. We further propose multiple instance learning (MIL) approaches at the level of images and points to supervise the object detector with point annotations. Finally, we propose a simple data augmentation, named Point-Guided Copy-Paste, to reduce the impact of those unmatched points. Experiments demonstrate the effectiveness of our method on a few datasets and various data regimes. In particular, Point-Teaching outperforms the previous best method Group R-CNN by 3.1 AP with 5% fully labeled data and 2.3 AP with 30% fully labeled data on the MS COCO dataset. We believe that our proposed framework can largely lower the bar of learning accurate object detectors and pave the way for its broader applications. The code is available at https://github.com/YongtaoGe/Point-Teaching.

Fast and accurate object detector for autonomous driving based on improved YOLOv5

Autonomous driving is an important branch of artificial intelligence, and real-time and accurate object detection is key to ensuring the safe and stable operation of autonomous vehicles. To this end, this paper proposes a fast and accurate object detector for autonomous driving based on improved YOLOv5. First, the YOLOv5 algorithm is improved by using structural re-parameterization (Rep), enhancing the accuracy and speed of the model through training-inference decoupling. Additionally, the neural architecture search method is introduced to cut redundant branches in the multi-branch re-parameterization module during the training phase, which ameliorates the training efficiency and accuracy. Finally, a small object detection layer is added to the network and the coordinate attention mechanism is added to all detection layers to improve the recognition rate of the model for small vehicles and pedestrians. The experimental results show that the detection accuracy of the proposed method on the KITTI dataset reaches 96.1%, and the FPS reaches 202, which is superior to many current mainstream algorithms and effectively improves the accuracy and real-time performance of unmanned driving object detection.

Toward surface defect detection in electronics manufacturing by an accurate and lightweight YOLO-style object detector

In electronics manufacturing, surface defect detection is very important for product quality control, and defective products can cause severe customer complaints. At the same time, in the manufacturing process, the cycle time of each product is usually very short. Furthermore, high-resolution input images from high-resolution industrial cameras are necessary to meet the requirements for high quality control standards. Hence, how to design an accurate object detector with real-time inference speed that can accept high-resolution input is an important task. In this work, an accurate YOLO-style object detector was designed, ATT-YOLO, which uses only one self-attention module, many-scale feature extraction and integration in the backbone and feature pyramid, and an improved auto-anchor design to address this problem. There are few datasets for surface detection in electronics manufacturing. Hence, we curated a dataset consisting of 14,478 laptop surface defects, on which ATT-YOLO achieved 92.8% mAP0.5 for the binary-class object detection task. We also further verified our design on the COCO benchmark dataset. Considering both computation costs and the performance of object detectors, ATT-YOLO outperforms several state-of-the-art and lightweight object detectors on the COCO dataset. It achieves a 44.9% mAP score and 21.8 GFLOPs, which is better than the compared models including YOLOv8-small (44.9%, 28.6G), YOLOv7-tiny-SiLU (38.7%, 13.8G), YOLOv6-small (43.1%, 44.2G), pp-YOLOE-small (42.7%, 17.4G), YOLOX-small (39.6%, 26.8G), and YOLOv5-small (36.7%, 17.2G). We hope that this work can serve as a useful reference for the utilization of attention-based networks in real-world situations.

Detection confidence driven multi-object tracking to recover reliable tracks from unreliable detections

Multi-object tracking (MOT) systems often rely on accurate object detectors; however, accurate detectors are not available in every application domain. We present Robust Confidence Tracking (RCT), an offline MOT algorithm designed for settings where detection quality is poor. Whereas prior methods simply threshold and discard detection confidence information, RCT relies on the exact detection confidence values to increase track quality throughout the entire tracking pipeline. This innovation (along with some simple and well-studied heuristics) allows RCT to achieve robust performance with minimal identity switches, even when provided with completely unfiltered detections. To compare trackers in the presence of unreliable detections, we present a challenging real-world underwater fish tracking dataset, FISHTRAC. In an large-scale evaluation across FISHTRAC, UA-DETRAC, and MOTChallenge data, RCT outperforms a wide variety of trackers, including deep trackers and more classic approaches. We have publically released our FISHTRAC codebase and training dataset at https://github.com/tmandel/fish-detrac, which will facilitate comparing trackers on understudied problems.

Improving apparel detection with category grouping and multi-grained branches

Training an accurate object detector is expensive and time-consuming. One main reason lies in the laborious labeling process, i.e., annotating category and bounding box information for all instances in every image. In this paper, we examine ways to improve performance of deep object detectors without extra labeling. We first explore to group existing categories of high visual and semantic similarities together as one super category (or, a superclass). Then, we study how this knowledge of hierarchical categories can be exploited to better detect objects using multi-grained RCNN branches∗. Experimental results on DeepFashion2 and OpenImagesV4-Clothing reveal that the proposed detection heads with multi-grained branches can boost the overall performance by as high as 2% with no additional time-consuming annotations. In addition, classes that have fewer training samples tend to benefit more from the proposed multi-grained heads with superclass grouping. In particular, we improve the mAP for the last 30% categories (in terms of training sample number) by 2.6% and 4.6% on DeepFashion2 and OpenImagesV4-Clothing, respectively.

Urban rail transit obstacle detection based on Improved R-CNN

Excellent active obstacle detection capability is critical to operate fully automatic trains safely and reliably. There are some problems exist in the traditional sensor-based obstacle detection approaches, such as low detection accuracy, sluggish detection speed and a limited number of obstacle types. In this work, a fast and accurate object detector termed improved R-CNN is proposed by introducing new up-sampling parallel structure and context extraction module (CEM) into the architecture of R-CNN. Furthermore, transfer learning is applied to inherit the COCO dataset's pre-training weight. The network is trained on track lines and test lines with nine types of obstacles. The data is evaluated and statistically cleansed, and the fine-tuning anchor improves the network's flexibility within the dataset. With the input size of 1330 px × 800 px, the test results show that the improved R-CNN model achieves an accuracy of 90.6% and a detection speed of 11 FPS. In comparison to other state-of-the-art detectors, the model has great performance in obstacle identification of rail track and achieves a good balance between detection speed and accuracy.

Learning Orientation Information From Frequency-Domain for Oriented Object Detection in Remote Sensing Images

Object detection in remote sensing images (RSIs) poses great difficulties due to arbitrary orientations, various scales and dense location of the targets over the ground. Recent evidence suggests that encoding the orientation information is of great use for training an accurate object detector for oriented object detection (OOD). In this paper, we propose a new frequency-domain orientation learning (FDOL) module with two main components: the frequency domain feature extraction (FFE) network and an orientation enhanced self-attention layer (OES-Layer). The FFE network models the interactions among spatial locations in the frequency domain to determine the frequency of spatial features. Then, these features are fed into our OES-Layer to learn the orientation information. Moreover, the orientation weights are adopted to guide the feature selection in a self-attention architecture, using them as a control gate to emphasize the spatial responses of target instances. Considering that the original similarity weights (calculated by the self-attention algorithm) do not distinctly model the orientation variation, the considered orientation weights provide an efficient asset to emphasize the orientation of objects. Extensive experiments on the DOTA and HRSC2016 datasets demonstrate that our method achieves state-of-the-art performance among single-scale methods, while achieving competitive performance over multi-scale methods.

SA-YOLOv3: An Efficient and Accurate Object Detector Using Self-Attention Mechanism for Autonomous Driving

Object detection is becoming increasingly significant for autonomous-driving system. However, poor accuracy or low inference performance limits current object detectors in applying to autonomous driving. In this work, a fast and accurate object detector termed as SA-YOLOv3, is proposed by introducing dilated convolution and self-attention module (SAM) into the architecture of YOLOv3. Furthermore, loss function based on GIoU and focal loss is reconstructed to further optimize detection performance. With an input size of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$512\times 512$ </tex-math></inline-formula> , our proposed SA-YOLOv3 improves YOLOv3 by 2.58 mAP and 2.63 mAP on KITTI and BDD100K benchmarks, with real-time inference (more than 40 FPS). When compared with other state-of-the-art detectors, it reports better trade-off in terms of detection accuracy and speed, indicating the suitability for autonomous-driving application. To our best knowledge, it is the first method that incorporates YOLOv3 with attention mechanism, and we expect this work would guide for autonomous-driving research in the future.

Active Player Detection in Handball Scenes Based on Activity Measures.

In team sports training scenes, it is common to have many players on the court, each with his own ball performing different actions. Our goal is to detect all players in the handball court and determine the most active player who performs the given handball technique. This is a very challenging task, for which, apart from an accurate object detector, which is able to deal with complex cluttered scenes, additional information is needed to determine the active player. We propose an active player detection method that combines the Yolo object detector, activity measures, and tracking methods to detect and track active players in time. Different ways of computing player activity were considered and three activity measures are proposed based on optical flow, spatiotemporal interest points, and convolutional neural networks. For tracking, we consider the use of the Hungarian assignment algorithm and the more complex Deep SORT tracker that uses additional visual appearance features to assist the assignment process. We have proposed the evaluation measure to evaluate the performance of the proposed active player detection method. The method is successfully tested on a custom handball video dataset that was acquired in the wild and on basketball video sequences. The results are commented on and some of the typical cases and issues are shown.

An Accurate Object Detector With Effective Feature Extraction by Intrinsic Prior Knowledge

Deep learning-based object detection algorithms play an increasingly important role in many computer vision tasks. Many detection algorithms are trying to increase the depth of networks to improve their feature expression capacity. However, how to better extract the essential features from a fixed number of training samples is often ignored. Thus, we propose CoReSh network, using the prior knowledge of intrinsic features, to address this problem. First, the relevant optical factors which may affect the detection performance are explored, and the intrinsic features are proven to be able to represent the essential information of objects. Second, a new data augmentation is proposed to increase the feature diversities of training samples through intrinsic decomposition. Third, a deep detection neural network incorporating color, reflectance, and shading image's information is designed. To optimize the performance of our network, different learning rate setting tactics are compared. Besides, a contrast experiment on the Pascal VOC dataset was set up to verify it. The result shows that our network is 10% better than ordinary deep learning-based detection networks in precision, and roughly the same as the network loaded with pre-training weights from ILSVRC. It shows that our network has greatly improved the precision of object detection and it is suitable when the training process lacks sufficient training samples.