Detection Benchmark Research Articles

CaLiJD: Camera and LiDAR Joint Contender for 3D Object Detection

Three-dimensional object detection has been a key area of research in recent years because of its rich spatial information and superior performance in addressing occlusion issues. However, the performance of 3D object detection still lags significantly behind that of 2D object detection, owing to challenges such as difficulties in feature extraction and a lack of texture information. To address this issue, this study proposes a 3D object detection network, CaLiJD (Camera and Lidar Joint Contender for 3D object Detection), guided by two-dimensional detection results. CaLiJD creatively integrates advanced channel attention mechanisms with a novel bounding-box filtering method to improve detection accuracy, especially for small and occluded objects. Bounding boxes are detected by the 2D and 3D networks for the same object in the same scene as an associated pair. The detection results that satisfy the criteria are then fed into the fusion layer for training. In this study, a novel fusion network is proposed. It consists of numerous convolutions arranged in both sequential and parallel forms and includes a Grouped Channel Attention Module for extracting interactions among multi-channel information. Moreover, a novel bounding-box filtering mechanism was introduced, incorporating the normalized distance from the object to the radar as a filtering criterion within the process. Experiments were conducted using the KITTI 3D object detection benchmark. The results showed that a substantial improvement in mean Average Precision (mAP) was achieved by CaLiJD compared with the baseline single-modal 3D detection model, with an enhancement of 7.54%. Moreover, the improvement achieved by our method surpasses that of other classical fusion networks by an additional 0.82%. In particular, CaLiJD achieved mAP values of 73.04% and 59.86%, respectively, thus demonstrating state-of-the-art performance for challenging small-object detection tasks such as those involving cyclists and pedestrians.

Self-similarity of communities of the ABCD model

The Artificial Benchmark for Community Detection (ABCD) graph is a random graph model with community structure and power-law distribution for both degrees and community sizes. The model generates graphs similar to the well-known LFR model but it is faster and can be investigated analytically. In this paper, we show that the ABCD model exhibits some interesting self-similar behaviour, namely, the degree distribution of ground-truth communities is asymptotically the same as the degree distribution of the whole graph (appropriately normalized based on their sizes). As a result, we can not only estimate the number of edges induced by each community but also the number of self-loops and multi-edges generated during the process. Understanding these quantities is important as (a) rewiring self-loops and multi-edges to keep the graph simple is an expensive part of the algorithm, and (b) every rewiring causes the underlying configuration models to deviate slightly from uniform simple graphs on their corresponding degree sequences.

CamoFormer: Masked Separable Attention for Camouflaged Object Detection.

How to identify and segment camouflaged objects from the background is challenging. Inspired by the multi-head self-attention in Transformers, we present a simple masked separable attention (MSA) for camouflaged object detection. We first separate the multi-head self-attention into three parts, which are responsible for distinguishing the camouflaged objects from the background using different mask strategies. Furthermore, we propose to capture high-resolution semantic representations progressively based on a simple top-down decoder with the proposed MSA to attain precise segmentation results. These structures plus a backbone encoder form a new model, dubbed CamoFormer. Extensive experiments show that CamoFormer achieves new state-of-the-art performance on three widely-used camouflaged object detection benchmarks. To better evaluate the performance of the proposed CamoFormer around the border regions, we propose to use two new metrics, i.e., BR-M and BR-F. There are on average ∼ 5% relative improvements over previous methods in terms of S-measure and weighted F-measure.

Learning to Holistically Detect Bridges From Large-Size VHR Remote Sensing Imagery.

Bridge detection in remote sensing images (RSIs) plays a crucial role in various applications, but it poses unique challenges compared to the detection of other objects. In RSIs, bridges exhibit considerable variations in terms of their spatial scales and aspect ratios. Therefore, to ensure the visibility and integrity of bridges, it is essential to perform holistic bridge detection in large-size very-high-resolution (VHR) RSIs. However, the lack of datasets with large-size VHR RSIs limits the deep learning algorithms' performance on bridge detection. Due to the limitation of GPU memory in tackling large-size images, deep learning-based object detection methods commonly adopt the cropping strategy, which inevitably results in label fragmentation and discontinuous prediction. To ameliorate the scarcity of datasets, this paper proposes a large-scale dataset named GLH-Bridge comprising 6,000 VHR RSIs sampled from diverse geographic locations across the globe. These images encompass a wide range of sizes, varying from 2,048 × 2,048 to 16,384 × 16,384 pixels, and collectively feature 59,737 bridges. These bridges span diverse backgrounds, and each of them has been manually annotated, using both an oriented bounding box (OBB) and a horizontal bounding box (HBB). Furthermore, we present an efficient network for holistic bridge detection (HBD-Net) in large-size RSIs. The HBD-Net presents a separate detector-based feature fusion (SDFF) architecture and is optimized via a shape-sensitive sample re-weighting (SSRW) strategy. The SDFF architecture performs inter-layer feature fusion (IFF) to incorporate multi-scale context in the dynamic image pyramid (DIP) of the large-size image, and the SSRW strategy is employed to ensure an equitable balance in the regression weight of bridges with various aspect ratios. Based on the proposed GLH-Bridge dataset, we establish a bridge detection benchmark including the OBB and HBB tasks, and validate the effectiveness of the proposed HBD-Net. Additionally, cross-dataset generalization experiments on two publicly available datasets illustrate the strong generalization capability of the GLH-Bridge dataset.

End to end polysemantic cooperative mixed task trainer for UAV target detection

With the rapid advancement and application of Unmanned Aerial Vehicles (UAVs), target detection in urban scenes has made significant progress. Achieving precise 3D reconstruction from oblique imagery is essential for accurate urban object detection in UAV images. However, challenges persist due to low detection accuracy caused by subtle target features, complex backgrounds, and the prevalence of small targets. To address these issues, we introduce the Polysemantic Cooperative Detection Transformer (Pc-DETR), a novel end-to-end UAV image target detection network. Our primary innovation, the Polysemantic Transformer (PoT) Backbone, enhances visual representation by leveraging contextual information to guide a dynamic attention matrix. This matrix, formed through convolutions, captures both static and dynamic features, resulting in superior detection. Additionally, we propose the Polysemantic Cooperative Mixed-Task Training scheme, which employs multiple auxiliary heads for diverse label assignments, boosting the encoder’s learning capacity. This approach customizes queries and optimizes training efficiency without increasing inference costs. Comparative experiments show that Pc-DETR achieves a 3% improvement in detection accuracy over the current state-of-the-art MFEFNet, setting a new benchmark in UAV image detection and advancing methodologies for intelligent UAV surveillance systems.

Anomaly Detection Using Complete Cycle Consistent Generative Adversarial Network.

This research presents a robust adversarial method for anomaly detection in real-world scenarios, leveraging the power of generative adversarial neural networks (GANs) through cycle consistency in reconstruction error. Traditional approaches often falter due to high variance in class-wise accuracy, rendering them ineffective across different anomaly types. Our proposed model addresses these challenges by introducing an innovative flow of information in the training procedure and integrating it as a new discriminator into the framework, thereby optimizing the training dynamics. Furthermore, it employs a supplementary distribution in the input space to steer reconstructions toward the normal data distribution. This adjustment distinctly isolates anomalous instances and enhances detection precision. Also, two unique anomaly scoring mechanisms were developed to augment detection capabilities. Comprehensive evaluations on six varied datasets have confirmed that our model outperforms one-class anomaly detection benchmarks. The implementation is openly accessible to the academic community, available on Github.a.

Optimizing Fire Scene Analysis: Hybrid Convolutional Neural Network Model Leveraging Multiscale Feature and Attention Mechanisms

The rapid and accurate detection of fire scenes in various environments is crucial for effective disaster management and mitigation. Fire scene classification is a critical aspect of modern fire detection systems that directly affects public safety and property preservation. This research introduced a novel hybrid deep learning model designed to enhance the accuracy and efficiency of fire scene classification across diverse environments. The proposed model integrates advanced convolutional neural networks with multiscale feature extraction, attention mechanisms, and ensemble learning to achieve superior performance in real-time fire detection. By leveraging the strengths of pre-trained networks such as ResNet50, VGG16, and EfficientNet-B3, the model captures detailed features at multiple scales, ensuring robust detection capabilities. Including spatial and channel attention mechanisms further refines the focus on critical areas within the input images, reducing false positives and improving detection precision. Extensive experiments on a comprehensive dataset encompassing wildfires, building fires, vehicle fires, and non-fire scenes demonstrate that the proposed framework outperforms existing cutting-edge techniques. The model also exhibited reduced computational complexity and enhanced inference speed, making it suitable for deployment in real-time applications on various hardware platforms. This study sets a new benchmark for fire detection and offers a powerful tool for early warning systems and emergency response initiatives.

BiFPN-YOLO: One-stage object detection integrating Bi-Directional Feature Pyramid Networks

Object detection is a key component in computer vision research, allowing a system to determine the location and type of object within any given scene. YOLOv5 is a modern object detection model, which utilises the advantages of the original YOLO implementation while being built from scratch in Python. In this paper, BiFPN-YOLO is proposed, featuring clear improvements over the existing range of YOLOv5 object detection models; these include replacing the traditional Path-Aggregation Network (PANet) with a higher performing Bi-Directional Feature Pyramid Network (BiFPN), requiring complex adaptation from its original implementation to function with YOLOv5, as well as exploring a replacement to the standard Swish activation function by evaluating the performance against a number of other activation functions. The proposed model showcases state-of-the-art performance, benchmarking against well-known datasets such as the German Traffic Sign Detection Benchmark (GTSDB), improving mAP by 3.1 %, and the RoboFEI@Home dataset, where Mean Average Precision (mAP) is improved by 2 % compared to the base YOLOv5 model. Performance was also improved on MSCOCO by 1.1 % and a custom subset of the OpenImagesV6 dataset by 2.4 %.

MSAttnFlow: Normalizing flow for unsupervised anomaly detection with multi-scale attention

Unsupervised anomaly detection (UAD) aims to locate anomalies in images without using annotated defective data. Normalizing flow is inherently suitable for the UAD task because it can explicitly model the data distribution and perform accurate density estimations. The multi-scale problem of anomalies is one of the key issues that existing flow-based methods focus on. To solve this problem, existing methods either introduce complex cross-scale information exchange operations or use multiple independent parallel flows. In this work, we propose a novel normalizing flow network with multi-scale attention, MSAttnFlow. We construct normalizing flows at different levels of the feature pyramid to handle multi-scale anomalies. Meanwhile, we introduce an autoregressive multi-scale attention module with two directions to fuse features of different scales, which facilitates the sharing of perception capabilities between different scales without introducing large computational overhead. MSAttnFlow achieves state-of-the-art on popular anomaly detection benchmarks MVTec-AD and VisA, with image/pixel level AUROC reaching 99.7/98.9 and 98.1/99.1 respectively. The performance of MSAttnFlow also improves by 2.9/1.3 on VisA compared to the previous SOTA method, which is a significant improvement considering the high performance of existing methods.

Enhancing Few-Shot Out-of-Distribution Detection with Pre-Trained Model Features.

Ensuring the reliability of open-world intelligent systems heavily relies on effective out-of-distribution (OOD) detection. Despite notable successes in existing OOD detection methods, their performance in scenarios with limited training samples is still suboptimal. Therefore, we first construct a comprehensive few-shot OOD detection benchmark in this paper. Remarkably, our investigation reveals that Parameter-Efficient Fine-Tuning (PEFT) techniques, such as visual prompt tuning and visual adapter tuning, outperform traditional methods like fully fine-tuning and linear probing tuning in few-shot OOD detection. Considering that some valuable information from the pre-trained model, which is conducive to OOD detection, may be lost during the fine-tuning process, we reutilize features from the pre-trained models to mitigate this issue. Specifically, we first propose a training-free approach, termed uncertainty score ensemble (USE). This method integrates feature-matching scores to enhance existing OOD detection methods, significantly narrowing the gap between traditional fine-tuning and PEFT techniques. However, due to its training-free property, this method is unable to improve in-distribution accuracy. To this end, we further propose a method called Domain-Specific and General Knowledge Fusion (DSGF) to improve few-shot OOD detection performance and ID accuracy under different fine-tuning paradigms. Experiment results demonstrate that DSGF enhances few-shot OOD detection across different fine-tuning strategies, shot settings, and OOD detection methods. We believe our work can provide the research community with a novel path to leveraging large-scale visual pre-trained models for addressing FS-OOD detection. The code will be released.

Towards superior android ransomware detection: An ensemble machine learning perspective

Ransomware remains a pervasive threat to Android devices, with its ability to encrypt critical data and demand ransoms causing significant disruptions to users and organizations alike. This research proposes a novel ensemble-based machine learning approach for the detection of Android ransomware, leveraging the strengths of multiple classifiers to enhance detection accuracy and robustness. Utilizing a comprehensive dataset comprising 203,556 network traffic records across 10 distinct ransomware types and benign traffic, we meticulously preprocess and feature-engineer the data to ensure optimal model performance. The methodology integrates various ensemble classifiers, evaluating each through rigorous cross-validation. Feature importance analysis using Random Forest identifies key indicators of ransomware activity, enabling us to refine our models and focus on the most predictive features. The results demonstrate that the ensemble models, particularly Bagging, achieve near-perfect detection rates, with precision, recall, and F1 scores consistently exceeding 99% for different binary attacks and multi-class classification. Finally, in-depth statistical analysis further validates the superiority of our approach, showcasing significant improvements over traditional machine learning methods. This research sets a new benchmark for Android ransomware detection, offering a robust, scalable, and highly accurate solution that enhances the security and resilience of mobile networks against evolving cyber threats.

Development and extensive sequencing of a broadly-consented Genome in a Bottle matched tumor-normal pair.

The Genome in a Bottle Consortium (GIAB), hosted by the National Institute of Standards and Technology (NIST), is developing new matched tumor-normal samples, the first to be explicitly consented for public dissemination of genomic data and cell lines. Here, we describe a comprehensive genomic dataset from the first individual, HG008, including DNA from an adherent, epithelial-like pancreatic ductal adenocarcinoma (PDAC) tumor cell line and matched normal cells from duodenal and pancreatic tissues. Data for the tumor-normal matched samples comes from thirteen distinct state-of-the-art whole genome measurement technologies, including high depth short and long-read bulk whole genome sequencing (WGS), single cell WGS, and Hi-C, and karyotyping. These data will be used by the GIAB Consortium to develop matched tumor-normal benchmarks for somatic variant detection. We expect these data to facilitate innovation for whole genome measurement technologies, de novo assembly of tumor and normal genomes, and bioinformatic tools to identify small and structural somatic mutations. This first-of-its-kind broadly consented open-access resource will facilitate further understanding of sequencing methods used for cancer biology.

Development of a High-Resolution Melting Method for the Detection of Clarithromycin-Resistant Helicobacter pylori in the Gastric Microbiome.

Background: The issue of Helicobacter pylori (H. pylori) resistance to clarithromycin (CLR) has consistently posed challenges for clinical treatment. Hence, a rapid susceptibility testing (AST) method urgently needs to be developed. Methods: In the present study, 35 isolates of H. pylori were isolated from 203 gastritis patients of the Guangzhou cohort, and the antimicrobial resistance phenotypes were associated with their genomes to analyze the relevant mutations. Based on these mutations, a rapid detection system utilizing high-resolution melting (HRM) curve analysis was designed and verified by the Shenzhen cohort, which consisted of 38 H. pylori strains. Results: Genomic analysis identified the mutation of the 2143 allele from A to G (A2143G) of 23S rRNA as the most relevant mutation with CLR resistance (p < 0.01). In the HRM system, the wild-type H. pylori showed a melting temperature (Tm) of 79.28 ± 0.01 °C, while the mutant type exhibited a Tm of 79.96 ± 0.01 °C. These differences enabled a rapid distinction between two types of H. pylori (p < 0.01). Verification examinations showed that this system could detect target DNA as low as 0.005 ng/μL in samples without being affected by other gastric microorganisms. The method also showed a good performance in the Shenzhen validation cohort, with 81.58% accuracy, and 100% specificity. Conclusions: We have developed an HRM system that can accurately and quickly detect CLR resistance in H. pylori. This method can be directly used for the detection of gastric microbiota samples and provides a new benchmark for the simple detection of H. pylori resistance.

A benchmark for Portuguese zero-shot stance detection

Stance detection is the task of inferring for/against attitudes towards a particular target from text. As targets are in principle unlimited, however, research in the field has moved from so-called in-domain classification (which assume the availability of a sufficient number of stances towards the intended target for training purposes) to more realistic zero-shot scenarios. However, regardless of which - or how much - training data is taken into account, most existing zero-shot approaches are devoted to the English language, in stark opposition to alternatives devoted to Portuguese. As a means to overcome some of these difficulties, this article presents a benchmark (hereby understood as the combination of a dataset, baseline systems and their results) for zero-shot Portuguese stance detection that is, to the best of our knowledge, the first of it kind. More specifically, we adapt a number of existing models available for the English language to Portuguese, and introduce novel approaches to the task based on more recent prompt engineering methods and off-task labelling, achieving SOTA results that are, in some cases, even superior to in-domain classification.

AMVFNet: Attentive Multi-View Fusion Network for 3D Object Detection

Pillar-based method is significant in the field of LiDAR-based 3D object detection which could directly make use of efficient 2D backbone and save computational resources during reference. Existing methods usually sequentially project the original point clouds into the cylindrical view or the bird-eye view for feature extraction. However, the former suffers from obscured problems and the scales of instances vary greatly with distance, while the latter leads to considerable confusion problems due to the loss of semantic information caused by the sparsity of the projected point cloud. In this paper, we present a novel and efficient two-stage point-pillar hybrid architecture named Attentive Multi-View Fusion Network (AMVFNet), in which we abstract features from all cylindrical view, bird-eye view, and raw point clouds. Rather than designing more complex modules to solve the problems inherent in the single-view approach, our multi-view fusion architecture effectively combines the strengths of multiple perspectives to improve performance at a more fundamental level. Besides, to compensate for quantization distortion caused by projection operations, we propose attentive feature enhancement layers to further improve the capability of contextual information capturing. Extensive experiments on the KITTI detection benchmark illustrate that our proposed AMVFNet achieves competitive performance compared with other SOTA 3D object detectors.

A hybrid CNN-LSTM approach for intelligent cyber intrusion detection system

As the technology is advancing more and more in the era of increasing digitalization, safeguarding networks from cyber threats is crucial. As cyber-attacks on critical infrastructure are becoming more and more sophisticated, enhancing cyber intrusion detection systems (IDS) is imperative. This paper proposes and evaluates a deep learning-based IDS using the NSL-KDD dataset, a benchmark for intrusion detection. The system pre-processes data with Recursive Feature Elimination (RFE) and a Decision Tree classifier to identify the most significant features, optimizing model performance. Various deep learning models, including ANN, LSTM, BiLSTM, CNN-LSTM, GRU, and BiGRU, have been evaluated. The CNN-LSTM model outperformed the others, with 95 % accuracy, 0.89 recall, and 0.94 f1-score. These results prove the effectiveness of the proposed IDS in accurately distinguishing between malicious and benign network traffic. Future research can explore ensemble techniques like boosting or bagging to further enhance IDS performance.

SSN: Scale Selection Network for Multi-Scale Object Detection in Remote Sensing Images

The rapid growth of deep learning technology has made object detection in remote sensing images an important aspect of computer vision, finding applications in military surveillance, maritime rescue, and environmental monitoring. Nonetheless, the capture of remote sensing images at high altitudes causes significant scale variations, resulting in a heterogeneous range of object scales. These varying scales pose significant challenges for detection algorithms. To solve the scale variation problem, traditional detection algorithms compute multi-layer feature maps. However, this approach introduces significant computational redundancy. Inspired by the mechanism of cognitive scaling mechanisms handling multi-scale information, we propose a novel Scale Selection Network (SSN) to eliminate computational redundancy through scale attentional allocation. In particular, we have devised a lightweight Landmark Guided Scale Attention Network, which is capable of predicting potential scales in an image. The detector only needs to focus on the selected scale features, which greatly reduces the inference time. Additionally, a fast Reversible Scale Semantic Flow Preserving strategy is proposed to directly generate multi-scale feature maps for detection. Experiments demonstrate that our method facilitates the acceleration of image pyramid-based detectors by approximately 5.3 times on widely utilized remote sensing object detection benchmarks.

STMixer: A One-Stage Sparse Action Detector.

Traditional video action detectors typically adopt the two-stage pipeline, where a person detector is first employed to generate actor boxes and then 3D RoIAlign is used to extract actor-specific features for action recognition. This detection paradigm requires multi-stage training and inference, and the feature sampling is only constrained inside the box, failing to effectively leverage richer context information outside. Recently, several query-based action detectors have been proposed to predict action instances in an end-to-end manner. However, they still lack adaptability in feature sampling and decoding, thus suffering from the issues of inferior performance or slower convergence. In this paper, we propose two core designs for a more flexible one-stage sparse action detector. First, we present a query-based adaptive feature sampling module, which endows the detector with the flexibility of mining a group of discriminative features from the entire spatio-temporal domain. Second, we devise a decoupled feature mixing module, which dynamically attends to and mixes video features along the spatial and temporal dimensions respectively for better feature decoding. Based on these designs, we instantiate two detection pipelines, that is, STMixer-K for keyframe action detection and STMixer-T for action tubelet detection. Without bells and whistles, our STMixer detectors obtain the state-of-the-art results on five challenging spatio-temporal action detection benchmarks for keyframe action detection or action tube detection.

Prospects for Revealing Intermediate-mass Black Holes in NGC 1399 Using SKA

This study investigates the detectability of intermediate-mass black holes (IMBHs) within the mass range 102 M ⊙ ≤ M BH ≤ 105 M ⊙ in the globular star clusters of NGC 1399 at a frequency of 300.00 MHz. Employing the theoretical Bondi accretion model and the empirical fundamental plane (FP) of black hole accretion, we estimate IMBH masses based on bolometric luminosity and X-ray/radio luminosities, respectively. By simulating a 3 hr observation of 77 globular cluster (GC) candidates using the Square Kilometre Array, we identify radio detection benchmarks indicative of accretion onto IMBHs. Our results show that IMBHs inside the globular star clusters located in NGC 1399 are indeed detectable, with the Bondi accretion model providing IMBH mass estimates ranging from 2.93 × 103.0±0.39 M ⊙ to 7.43 × 104.0±0.39 M ⊙ and the empirical FP relation suggesting IMBH mass estimation with 3.41 × 105.0±0.96 M ⊙. These findings highlight the presence and detectability of IMBHs in GCs, offering insights into their role as precursors to supermassive black holes and enriching our understanding of black hole formation and evolution in astrophysical environments.

Noise-robust modal parameter identification and damage assessment for aero-structures

PurposeGround vibration testing is critical for aircraft design and certification. Fast relaxed vector fitting (FRVF) and Loewner framework (LF), recently extended to modal parameter extraction in mechanical systems to address the computational challenges of time and frequency domain techniques, are applied for damage detection on aeronautically relevant structures.Design/methodology/approachFRVF and LF are applied to numerical datasets to assess noise robustness and performance for damage detection. Computational efficiency is also evaluated. In addition, they are applied to a novel damage detection benchmark of a high aspect ratio wing, comparing their performance with the state-of-the-art method N4SID.FindingsFRVF and LF detect structural changes effectively; LF exhibits better noise robustness, while FRVF is more computationally efficient.Practical implicationsLF is recommended for noisy measurements.Originality/valueTo the best of the authors’ knowledge, this is the first study in which the LF and FRVF are applied for the extraction of the modal parameters in aeronautically relevant structures. In addition, a novel damage detection benchmark of a high-aspect-ratio wing is introduced.