Abstract
The early and accurate diagnosis of sepsis is critical for enhancing patient outcomes. This study aims to use heart rate variability (HRV) features to develop an effective predictive model for sepsis detection. Critical HRV features are identified through feature engineering methods, including statistical bootstrapping and the Boruta algorithm, after which XGBoost and Random Forest classifiers are trained with differential hyperparameter settings. In addition, ensemble models are constructed to pool the prediction probabilities of high-recall and high-precision classifiers and improve model performance. Finally, a neural network model is trained on the HRV features, achieving an F1 score of 0.805, a precision of 0.851, and a recall of 0.763. The best-performing machine learning model is compared to this neural network through an interpretability analysis, where Local Interpretable Model-agnostic Explanations are implemented to determine decision-making criterion based on numerical ranges and thresholds for specific features. This study not only highlights the efficacy of HRV in automated sepsis diagnosis but also increases the transparency of black box outputs, maximizing clinical applicability.
Title:
Privacy-Safe Iris Presentation Attack Detection
Authors: Mahsa Mitcheff, Patrick Tinsley, Adam Czajka
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
Abstract
This paper proposes a framework for a privacy-safe iris presentation attack detection (PAD) method, designed solely with synthetically-generated, identity-leakage-free iris images. Once trained, the method is evaluated in a classical way using state-of-the-art iris PAD benchmarks. We designed two generative models for the synthesis of ISO/IEC 19794-6-compliant iris images. The first model synthesizes bona fide-looking samples. To avoid ``identity leakage,'' the generated samples that accidentally matched those used in the model's training were excluded. The second model synthesizes images of irises with textured contact lenses and is conditioned by a given contact lens brand to have better control over textured contact lens appearance when forming the training set. Our experiments demonstrate that models trained solely on synthetic data achieve a lower but still reasonable performance when compared to solutions trained with iris images collected from human subjects. This is the first-of-its-kind attempt to use solely synthetic data to train a fully-functional iris PAD solution, and despite the performance gap between regular and the proposed methods, this study demonstrates that with the increasing fidelity of generative models, creating such privacy-safe iris PAD methods may be possible. The source codes and generative models trained for this work are offered along with the paper.
Title:
A Novel Hybrid Approach for Tornado Prediction in the United States: Kalman-Convolutional BiLSTM with Multi-Head Attention
Abstract
Tornadoes are among the most intense atmospheric vortex phenomena and pose significant challenges for detection and forecasting. Conventional methods, which heavily depend on ground-based observations and radar data, are limited by issues such as decreased accuracy over greater distances and a high rate of false positives. To address these challenges, this study utilizes the Seamless Hybrid Scan Reflectivity (SHSR) dataset from the Multi-Radar Multi-Sensor (MRMS) system, which integrates data from multiple radar sources to enhance accuracy. A novel hybrid model, the Kalman-Convolutional BiLSTM with Multi-Head Attention, is introduced to improve dynamic state estimation and capture both spatial and temporal dependencies within the data. This model demonstrates superior performance in precision, recall, F1-Score, and accuracy compared to methods such as K-Nearest Neighbors (KNN) and LightGBM. The results highlight the considerable potential of advanced machine learning techniques to improve tornado prediction and reduce false alarm rates. Future research will focus on expanding datasets, exploring innovative model architectures, and incorporating large language models (LLMs) to provide deeper insights. This research introduces a novel model for tornado prediction, offering a robust framework for enhancing forecasting accuracy and public safety.
Title:
Refined Infrared Small Target Detection Scheme with Single-Point Supervision
Authors: Jinmiao Zhao, Zelin Shi, Chuang Yu, Yunpeng Liu
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Recently, infrared small target detection with single-point supervision has attracted extensive attention. However, the detection accuracy of existing methods has difficulty meeting actual needs. Therefore, we propose an innovative refined infrared small target detection scheme with single-point supervision, which has excellent segmentation accuracy and detection rate. Specifically, we introduce label evolution with single point supervision (LESPS) framework and explore the performance of various excellent infrared small target detection networks based on this framework. Meanwhile, to improve the comprehensive performance, we construct a complete post-processing strategy. On the one hand, to improve the segmentation accuracy, we use a combination of test-time augmentation (TTA) and conditional random field (CRF) for post-processing. On the other hand, to improve the detection rate, we introduce an adjustable sensitivity (AS) strategy for post-processing, which fully considers the advantages of multiple detection results and reasonably adds some areas with low confidence to the fine segmentation image in the form of centroid points. In addition, to further improve the performance and explore the characteristics of this task, on the one hand, we construct and find that a multi-stage loss is helpful for fine-grained detection. On the other hand, we find that a reasonable sliding window cropping strategy for test samples has better performance for actual multi-size samples. Extensive experimental results show that the proposed scheme achieves state-of-the-art (SOTA) performance. Notably, the proposed scheme won the third place in the "ICPR 2024 Resource-Limited Infrared Small Target Detection Challenge Track 1: Weakly Supervised Infrared Small Target Detection".
Title:
LR-Net: A Lightweight and Robust Network for Infrared Small Target Detection
Authors: Chuang Yu, Yunpeng Liu, Jinmiao Zhao, Zelin Shi
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Limited by equipment limitations and the lack of target intrinsic features, existing infrared small target detection methods have difficulty meeting actual comprehensive performance requirements. Therefore, we propose an innovative lightweight and robust network (LR-Net), which abandons the complex structure and achieves an effective balance between detection accuracy and resource consumption. Specifically, to ensure the lightweight and robustness, on the one hand, we construct a lightweight feature extraction attention (LFEA) module, which can fully extract target features and strengthen information interaction across channels. On the other hand, we construct a simple refined feature transfer (RFT) module. Compared with direct cross-layer connections, the RFT module can improve the network's feature refinement extraction capability with little resource consumption. Meanwhile, to solve the problem of small target loss in high-level feature maps, on the one hand, we propose a low-level feature distribution (LFD) strategy to use low-level features to supplement the information of high-level features. On the other hand, we introduce an efficient simplified bilinear interpolation attention module (SBAM) to promote the guidance constraints of low-level features on high-level features and the fusion of the two. In addition, We abandon the traditional resizing method and adopt a new training and inference cropping strategy, which is more robust to datasets with multi-scale samples. Extensive experimental results show that our LR-Net achieves state-of-the-art (SOTA) performance. Notably, on the basis of the proposed LR-Net, we achieve 3rd place in the "ICPR 2024 Resource-Limited Infrared Small Target Detection Challenge Track 2: Lightweight Infrared Small Target Detection".
Title:
Interpretation of the Intent Detection Problem as Dynamics in a Low-dimensional Space
Authors: Eduardo Sanchez-Karhunen, Jose F. Quesada-Moreno, Miguel A. Gutiérrez-Naranjo
Subjects: Subjects:
Machine Learning (cs.LG); Computation and Language (cs.CL); Machine Learning (stat.ML)
Abstract
Intent detection is a text classification task whose aim is to recognize and label the semantics behind a users query. It plays a critical role in various business applications. The output of the intent detection module strongly conditions the behavior of the whole system. This sequence analysis task is mainly tackled using deep learning techniques. Despite the widespread use of these techniques, the internal mechanisms used by networks to solve the problem are poorly understood. Recent lines of work have analyzed the computational mechanisms learned by RNNs from a dynamical systems perspective. In this work, we investigate how different RNN architectures solve the SNIPS intent detection problem. Sentences injected into trained networks can be interpreted as trajectories traversing a hidden state space. This space is constrained to a low-dimensional manifold whose dimensionality is related to the embedding and hidden layer sizes. To generate predictions, RNN steers the trajectories towards concrete regions, spatially aligned with the output layer matrix rows directions. Underlying the system dynamics, an unexpected fixed point topology has been identified with a limited number of attractors. Our results provide new insights into the inner workings of networks that solve the intent detection task.
Title:
Diverse Generation while Maintaining Semantic Coordination: A Diffusion-Based Data Augmentation Method for Object Detection
Authors: Sen Nie, Zhuo Wang, Xinxin Wang, Kun He
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Recent studies emphasize the crucial role of data augmentation in enhancing the performance of object detection models. However,existing methodologies often struggle to effectively harmonize dataset diversity with semantic this http URL bridge this gap, we introduce an innovative augmentation technique leveraging pre-trained conditional diffusion models to mediate this balance. Our approach encompasses the development of a Category Affinity Matrix, meticulously designed to enhance dataset diversity, and a Surrounding Region Alignment strategy, which ensures the preservation of semantic coordination in the augmented images. Extensive experimental evaluations confirm the efficacy of our method in enriching dataset diversity while seamlessly maintaining semantic coordination. Our method yields substantial average improvements of +1.4AP, +0.9AP, and +3.4AP over existing alternatives on three distinct object detection models, respectively.
Title:
Lighthouse: A User-Friendly Library for Reproducible Video Moment Retrieval and Highlight Detection
Abstract
We propose Lighthouse, a user-friendly library for reproducible video moment retrieval and highlight detection (MR-HD). Although researchers proposed various MR-HD approaches, the research community holds two main issues. The first is a lack of comprehensive and reproducible experiments across various methods, datasets, and video-text features. This is because no unified training and evaluation codebase covers multiple settings. The second is user-unfriendly design. Because previous works use different libraries, researchers set up individual environments. In addition, most works release only the training codes, requiring users to implement the whole inference process of MR-HD. Lighthouse addresses these issues by implementing a unified reproducible codebase that includes six models, three features, and five datasets. In addition, it provides an inference API and web demo to make these methods easily accessible for researchers and developers. Our experiments demonstrate that Lighthouse generally reproduces the reported scores in the reference papers. The code is available at this https URL.
Title:
Enabling Intelligent Traffic Systems: A Deep Learning Method for Accurate Arabic License Plate Recognition
Abstract
This paper introduces a novel two-stage framework for accurate Egyptian Vehicle License Plate Recognition (EVLPR). The first stage employs image processing techniques to reliably localize license plates, while the second stage utilizes a custom-designed deep learning model for robust Arabic character recognition. The proposed system achieves a remarkable 99.3% accuracy on a diverse dataset, surpassing existing approaches. Its potential applications extend to intelligent traffic management, including traffic violation detection and parking optimization. Future research will focus on enhancing the system's capabilities through architectural refinements, expanded datasets, and addressing system dependencies.
Title:
WWW: Where, Which and Whatever Enhancing Interpretability in Multimodal Deepfake Detection
Authors: Juho Jung, Sangyoun Lee, Jooeon Kang, Yunjin Na
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
All current benchmarks for multimodal deepfake detection manipulate entire frames using various generation techniques, resulting in oversaturated detection accuracies exceeding 94% at the video-level classification. However, these benchmarks struggle to detect dynamic deepfake attacks with challenging frame-by-frame alterations presented in real-world scenarios. To address this limitation, we introduce FakeMix, a novel clip-level evaluation benchmark aimed at identifying manipulated segments within both video and audio, providing insight into the origins of deepfakes. Furthermore, we propose novel evaluation metrics, Temporal Accuracy (TA) and Frame-wise Discrimination Metric (FDM), to assess the robustness of deepfake detection models. Evaluating state-of-the-art models against diverse deepfake benchmarks, particularly FakeMix, demonstrates the effectiveness of our approach comprehensively. Specifically, while achieving an Average Precision (AP) of 94.2% at the video-level, the evaluation of the existing models at the clip-level using the proposed metrics, TA and FDM, yielded sharp declines in accuracy to 53.1%, and 52.1%, respectively.
Title:
Enhancing Stability and Assessing Uncertainty in Community Detection through a Consensus-based Approach
Authors: Fabio Morea, Domenico De Stefano
Subjects: Subjects:
Social and Information Networks (cs.SI); Applications (stat.AP)
Abstract
Complex data in social and natural sciences find effective representation through networks, wherein quantitative and categorical information can be associated with nodes and connecting edges. The internal structure of networks can be explored using unsupervised machine learning methods known as community detection algorithms. The process of community detection is inherently subject to uncertainty as algorithms utilize heuristic approaches and randomised procedures to explore vast solution spaces, resulting in non-deterministic outcomes and variability in detected communities across multiple runs. Moreover, many algorithms are not designed to identify outliers and may fail to take into account that a network is an unordered mathematical entity. The main aim of our work is to address these issues through a consensus-based approach by introducing a new framework called Consensus Community Detection (CCD). Our method can be applied to different community detection algorithms, allowing the quantification of uncertainty for the whole network as well as for each node, and providing three strategies for dealing with outliers: incorporate, highlight, or group. The effectiveness of our approach is evaluated on artificial benchmark networks.
Title:
Automatic String Data Validation with Pattern Discovery
Abstract
In enterprise data pipelines, data insertions occur periodically and may impact downstream services if data quality issues are not addressed. Typically, such problems can be investigated and fixed by on-call engineers, but locating the cause of such problems and fixing errors are often time-consuming. Therefore, automatic data validation is a better solution to defend the system and downstream services by enabling early detection of errors and providing detailed error messages for quick resolution. This paper proposes a self-validate data management system with automatic pattern discovery techniques to verify the correctness of semi-structural string data in enterprise data pipelines. Our solution extracts patterns from historical data and detects erroneous incoming data in a top-down fashion. High-level information of historical data is analyzed to discover the format skeleton of correct values. Fine-grained semantic patterns are then extracted to strike a balance between generalization and specification of the discovered pattern, thus covering as many correct values as possible while avoiding over-fitting. To tackle cold start and rapid data growth, we propose an incremental update strategy and example generalization strategy. Experiments on large-scale industrial and public datasets demonstrate the effectiveness and efficiency of our method compared to alternative solutions. Furthermore, a case study on an industrial platform (Ant Group Inc.) with thousands of applications shows that our system captures meaningful data patterns in daily operations and helps engineers quickly identify errors.
Title:
CKNN: Cleansed k-Nearest Neighbor for Unsupervised Video Anomaly Detection
Authors: Jihun Yi, Sungroh Yoon
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
In this paper, we address the problem of unsupervised video anomaly detection (UVAD). The task aims to detect abnormal events in test video using unlabeled videos as training data. The presence of anomalies in the training data poses a significant challenge in this task, particularly because they form clusters in the feature space. We refer to this property as the "Anomaly Cluster" issue. The condensed nature of these anomalies makes it difficult to distinguish between normal and abnormal data in the training set. Consequently, training conventional anomaly detection techniques using an unlabeled dataset often leads to sub-optimal results. To tackle this difficulty, we propose a new method called Cleansed k-Nearest Neighbor (CKNN), which explicitly filters out the Anomaly Clusters by cleansing the training dataset. Following the k-nearest neighbor algorithm in the feature space provides powerful anomaly detection capability. Although the identified Anomaly Cluster issue presents a significant challenge to applying k-nearest neighbor in UVAD, our proposed cleansing scheme effectively addresses this problem. We evaluate the proposed method on various benchmark datasets and demonstrate that CKNN outperforms the previous state-of-the-art UVAD method by up to 8.5% (from 82.0 to 89.0) in terms of AUROC. Moreover, we emphasize that the performance of the proposed method is comparable to that of the state-of-the-art method trained using anomaly-free data.
Title:
Nighttime Pedestrian Detection Based on Fore-Background Contrast Learning
Authors: He Yao, Yongjun Zhang, Huachun Jian, Li Zhang, Ruzhong Cheng
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
The significance of background information is frequently overlooked in contemporary research concerning channel attention mechanisms. This study addresses the issue of suboptimal single-spectral nighttime pedestrian detection performance under low-light conditions by incorporating background information into the channel attention mechanism. Despite numerous studies focusing on the development of efficient channel attention mechanisms, the relevance of background information has been largely disregarded. By adopting a contrast learning approach, we reexamine channel attention with regard to pedestrian objects and background information for nighttime pedestrian detection, resulting in the proposed Fore-Background Contrast Attention (FBCA). FBCA possesses two primary attributes: (1) channel descriptors form remote dependencies with global spatial feature information; (2) the integration of background information enhances the distinction between channels concentrating on low-light pedestrian features and those focusing on background information. Consequently, the acquired channel descriptors exhibit a higher semantic level and spatial accuracy. Experimental outcomes demonstrate that FBCA significantly outperforms existing methods in single-spectral nighttime pedestrian detection, achieving state-of-the-art results on the NightOwls and TJU-DHD-pedestrian datasets. Furthermore, this methodology also yields performance improvements for the multispectral LLVIP dataset. These findings indicate that integrating background information into the channel attention mechanism effectively mitigates detector performance degradation caused by illumination factors in nighttime scenarios.
Title:
OpenOmni: A Collaborative Open Source Tool for Building Future-Ready Multimodal Conversational Agents
Abstract
Multimodal conversational agents are highly desirable because they offer natural and human-like interaction. However, there is a lack of comprehensive end-to-end solutions to support collaborative development and benchmarking. While proprietary systems like GPT-4o and Gemini demonstrating impressive integration of audio, video, and text with response times of 200-250ms, challenges remain in balancing latency, accuracy, cost, and data privacy. To better understand and quantify these issues, we developed OpenOmni, an open-source, end-to-end pipeline benchmarking tool that integrates advanced technologies such as Speech-to-Text, Emotion Detection, Retrieval Augmented Generation, Large Language Models, along with the ability to integrate customized models. OpenOmni supports local and cloud deployment, ensuring data privacy and supporting latency and accuracy benchmarking. This flexible framework allows researchers to customize the pipeline, focusing on real bottlenecks and facilitating rapid proof-of-concept development. OpenOmni can significantly enhance applications like indoor assistance for visually impaired individuals, advancing human-computer interaction. Our demonstration video is available this https URL, demo is available via this https URL, code is available via this https URL.
Title:
Enhancing Twitter Bot Detection via Multimodal Invariant Representations
Authors: Jibing Gong, Jiquan Peng, Jin Qu, ShuYing Du, Kaiyu Wang
Subjects: Subjects:
Social and Information Networks (cs.SI)
Abstract
Detecting Twitter Bots is crucial for maintaining the integrity of online discourse, safeguarding democratic processes, and preventing the spread of malicious propaganda. However, advanced Twitter Bots today often employ sophisticated feature manipulation and account farming techniques to blend seamlessly with genuine user interactions, posing significant challenges to existing detection models. In response to these challenges, this paper proposes a novel Twitter Bot Detection framework called BotSAI. This framework enhances the consistency of multimodal user features, accurately characterizing various modalities to distinguish between real users and bots. Specifically, the architecture integrates information from users, textual content, and heterogeneous network topologies, leveraging customized encoders to obtain comprehensive user feature representations. The heterogeneous network encoder efficiently aggregates information from neighboring nodes through oversampling techniques and local relationship transformers. Subsequently, a multi-channel representation mechanism maps user representations into invariant and specific subspaces, enhancing the feature vectors. Finally, a self-attention mechanism is introduced to integrate and refine the enhanced user representations, enabling efficient information interaction. Extensive experiments demonstrate that BotSAI outperforms existing state-of-the-art methods on two major Twitter Bot Detection benchmarks, exhibiting superior performance. Additionally, systematic experiments reveal the impact of different social relationships on detection accuracy, providing novel insights for the identification of social bots.
Title:
Automated Defects Detection and Fix in Logging Statement
Authors: Renyi Zhong, Yichen Li, Jinxi Kuang, Wenwei Gu, Yintong Huo, Michael R. Lyu
Abstract
Developers use logging statements to monitor software, but misleading logs can complicate maintenance by obscuring actual activities. Existing research on logging quality issues is limited, mainly focusing on single defects and manual fixes. To address this, we conducted a study identifying four defect types in logging statements through real-world log changes analysis. We propose LogFixer, a two-stage framework for automatic detection and updating of logging statements. In the offline stage, LogFixer uses a similarity-based classifier on synthetic defective logs to identify defects. During the online phase, this classifier evaluates logs in code snippets to determine necessary improvements, and an LLM-based recommendation framework suggests updates based on historical log changes. We evaluated LogFixer on real-world and synthetic datasets, and new real-world projects, achieving an F1 score of 0.625. LogFixer significantly improved static text and dynamic variables suggestions by 48.12\% and 24.90\%, respectively, and achieved a 61.49\% success rate in recommending correct updates for new projects. We reported 40 problematic logs to GitHub, resulting in 25 confirmed and merged changes across 11 projects.
Title:
SuperSimpleNet: Unifying Unsupervised and Supervised Learning for Fast and Reliable Surface Defect Detection
Authors: Blaž Rolih, Matic Fučka, Danijel Skočaj
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
The aim of surface defect detection is to identify and localise abnormal regions on the surfaces of captured objects, a task that's increasingly demanded across various industries. Current approaches frequently fail to fulfil the extensive demands of these industries, which encompass high performance, consistency, and fast operation, along with the capacity to leverage the entirety of the available training data. Addressing these gaps, we introduce SuperSimpleNet, an innovative discriminative model that evolved from SimpleNet. This advanced model significantly enhances its predecessor's training consistency, inference time, as well as detection performance. SuperSimpleNet operates in an unsupervised manner using only normal training images but also benefits from labelled abnormal training images when they are available. SuperSimpleNet achieves state-of-the-art results in both the supervised and the unsupervised settings, as demonstrated by experiments across four challenging benchmark datasets. Code: this https URL .
Title:
Malicious Internet Entity Detection Using Local Graph Inference
Authors: Simon Mandlik, Tomas Pevny, Vaclav Smidl, Lukas Bajer
Subjects: Subjects:
Machine Learning (cs.LG); Cryptography and Security (cs.CR)
Abstract
Detection of malicious behavior in a large network is a challenging problem for machine learning in computer security, since it requires a model with high expressive power and scalable inference. Existing solutions struggle to achieve this feat -- current cybersec-tailored approaches are still limited in expressivity, and methods successful in other domains do not scale well for large volumes of data, rendering frequent retraining impossible. This work proposes a new perspective for learning from graph data that is modeling network entity interactions as a large heterogeneous graph. High expressivity of the method is achieved with neural network architecture HMILnet that naturally models this type of data and provides theoretical guarantees. The scalability is achieved by pursuing local graph inference, i.e., classifying individual vertices and their neighborhood as independent samples. Our experiments exhibit improvement over the state-of-the-art Probabilistic Threat Propagation (PTP) algorithm, show a further threefold accuracy improvement when additional data is used, which is not possible with the PTP algorithm, and demonstrate the generalization capabilities of the method to new, previously unseen entities.
Title:
Training LLMs to Recognize Hedges in Spontaneous Narratives
Authors: Amie J. Paige, Adil Soubki, John Murzaku, Owen Rambow, Susan E. Brennan
Subjects: Subjects:
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Abstract
Hedges allow speakers to mark utterances as provisional, whether to signal non-prototypicality or "fuzziness", to indicate a lack of commitment to an utterance, to attribute responsibility for a statement to someone else, to invite input from a partner, or to soften critical feedback in the service of face-management needs. Here we focus on hedges in an experimentally parameterized corpus of 63 Roadrunner cartoon narratives spontaneously produced from memory by 21 speakers for co-present addressees, transcribed to text (Galati and Brennan, 2010). We created a gold standard of hedges annotated by human coders (the Roadrunner-Hedge corpus) and compared three LLM-based approaches for hedge detection: fine-tuning BERT, and zero and few-shot prompting with GPT-4o and LLaMA-3. The best-performing approach was a fine-tuned BERT model, followed by few-shot GPT-4o. After an error analysis on the top performing approaches, we used an LLM-in-the-Loop approach to improve the gold standard coding, as well as to highlight cases in which hedges are ambiguous in linguistically interesting ways that will guide future research. This is the first step in our research program to train LLMs to interpret and generate collateral signals appropriately and meaningfully in conversation.
Keyword: face recognition
There is no result
Keyword: augmentation
Title:
Symmetric Graph Contrastive Learning against Noisy Views for Recommendation
Abstract
Graph Contrastive Learning (GCL) leverages data augmentation techniques to produce contrasting views, enhancing the accuracy of recommendation systems through learning the consistency between contrastive views. However, existing augmentation methods, such as directly perturbing interaction graph (e.g., node/edge dropout), may interfere with the original connections and generate poor contrasting views, resulting in sub-optimal performance. In this paper, we define the views that share only a small amount of information with the original graph due to poor data augmentation as noisy views (i.e., the last 20% of the views with a cosine similarity value less than 0.1 to the original view). We demonstrate through detailed experiments that noisy views will significantly degrade recommendation performance. Further, we propose a model-agnostic Symmetric Graph Contrastive Learning (SGCL) method with theoretical guarantees to address this issue. Specifically, we introduce symmetry theory into graph contrastive learning, based on which we propose a symmetric form and contrast loss resistant to noisy interference. We provide theoretical proof that our proposed SGCL method has a high tolerance to noisy views. Further demonstration is given by conducting extensive experiments on three real-world datasets. The experimental results demonstrate that our approach substantially increases recommendation accuracy, with relative improvements reaching as high as 12.25% over nine other competing models. These results highlight the efficacy of our method.
Title:
RCDM: Enabling Robustness for Conditional Diffusion Model
Abstract
The conditional diffusion model (CDM) enhances the standard diffusion model by providing more control, improving the quality and relevance of the outputs, and making the model adaptable to a wider range of complex tasks. However, inaccurate conditional inputs in the inverse process of CDM can easily lead to generating fixed errors in the neural network, which diminishes the adaptability of a well-trained model. The existing methods like data augmentation, adversarial training, robust optimization can improve the robustness, while they often face challenges such as high computational complexity, limited applicability to unknown perturbations, and increased training difficulty. In this paper, we propose a lightweight solution, the Robust Conditional Diffusion Model (RCDM), based on control theory to dynamically reduce the impact of noise and significantly enhance the model's robustness. RCDM leverages the collaborative interaction between two neural networks, along with optimal control strategies derived from control theory, to optimize the weights of two networks during the sampling process. Unlike conventional techniques, RCDM establishes a mathematical relationship between fixed errors and the weights of the two neural networks without incurring additional computational overhead. Extensive experiments were conducted on MNIST and CIFAR-10 datasets, and the results demonstrate the effectiveness and adaptability of our proposed model.
Title:
Refined Infrared Small Target Detection Scheme with Single-Point Supervision
Authors: Jinmiao Zhao, Zelin Shi, Chuang Yu, Yunpeng Liu
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Recently, infrared small target detection with single-point supervision has attracted extensive attention. However, the detection accuracy of existing methods has difficulty meeting actual needs. Therefore, we propose an innovative refined infrared small target detection scheme with single-point supervision, which has excellent segmentation accuracy and detection rate. Specifically, we introduce label evolution with single point supervision (LESPS) framework and explore the performance of various excellent infrared small target detection networks based on this framework. Meanwhile, to improve the comprehensive performance, we construct a complete post-processing strategy. On the one hand, to improve the segmentation accuracy, we use a combination of test-time augmentation (TTA) and conditional random field (CRF) for post-processing. On the other hand, to improve the detection rate, we introduce an adjustable sensitivity (AS) strategy for post-processing, which fully considers the advantages of multiple detection results and reasonably adds some areas with low confidence to the fine segmentation image in the form of centroid points. In addition, to further improve the performance and explore the characteristics of this task, on the one hand, we construct and find that a multi-stage loss is helpful for fine-grained detection. On the other hand, we find that a reasonable sliding window cropping strategy for test samples has better performance for actual multi-size samples. Extensive experimental results show that the proposed scheme achieves state-of-the-art (SOTA) performance. Notably, the proposed scheme won the third place in the "ICPR 2024 Resource-Limited Infrared Small Target Detection Challenge Track 1: Weakly Supervised Infrared Small Target Detection".
Title:
Diverse Generation while Maintaining Semantic Coordination: A Diffusion-Based Data Augmentation Method for Object Detection
Authors: Sen Nie, Zhuo Wang, Xinxin Wang, Kun He
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Recent studies emphasize the crucial role of data augmentation in enhancing the performance of object detection models. However,existing methodologies often struggle to effectively harmonize dataset diversity with semantic this http URL bridge this gap, we introduce an innovative augmentation technique leveraging pre-trained conditional diffusion models to mediate this balance. Our approach encompasses the development of a Category Affinity Matrix, meticulously designed to enhance dataset diversity, and a Surrounding Region Alignment strategy, which ensures the preservation of semantic coordination in the augmented images. Extensive experimental evaluations confirm the efficacy of our method in enriching dataset diversity while seamlessly maintaining semantic coordination. Our method yields substantial average improvements of +1.4AP, +0.9AP, and +3.4AP over existing alternatives on three distinct object detection models, respectively.
Keyword: detection
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Keyword: face recognition
There is no result
Keyword: augmentation
Title:
Title:
Title:
Title: