Abstract
This paper proposes a novel framework for audio deepfake detection with two main objectives: i) attaining the highest possible accuracy on available fake data, and ii) effectively performing continuous learning on new fake data in a few-shot learning manner. Specifically, we conduct a large audio deepfake collection using various deep audio generation methods. The data is further enhanced with additional augmentation methods to increase variations amidst compressions, far-field recordings, noise, and other distortions. We then adopt the Audio Spectrogram Transformer for the audio deepfake detection model. Accordingly, the proposed method achieves promising performance on various benchmark datasets. Furthermore, we present a continuous learning plugin module to update the trained model most effectively with the fewest possible labeled data points of the new fake type. The proposed method outperforms the conventional direct fine-tuning approach with much fewer labeled data points.
Title:
Advance and Refinement: The Evolution of UAV Detection and Classification Technologies
Abstract
This review provides a detailed analysis of the advancements in unmanned aerial vehicle (UAV) detection and classification systems from 2020 to today. It covers various detection methodologies such as radar, radio frequency, optical, and acoustic sensors, and emphasizes their integration via sophisticated sensor fusion techniques. The fundamental technologies driving UAV detection and classification are thoroughly examined, with a focus on their accuracy and range. Additionally, the paper discusses the latest innovations in artificial intelligence and machine learning, illustrating their impact on improving the accuracy and efficiency of these systems. The review concludes by predicting further technological developments in UAV detection, which are expected to enhance both performance and reliability.
Title:
Investigating Causal Cues: Strengthening Spoofed Audio Detection with Human-Discernible Linguistic Features
Abstract
Several types of spoofed audio, such as mimicry, replay attacks, and deepfakes, have created societal challenges to information integrity. Recently, researchers have worked with sociolinguistics experts to label spoofed audio samples with Expert Defined Linguistic Features (EDLFs) that can be discerned by the human ear: pitch, pause, word-initial and word-final release bursts of consonant stops, audible intake or outtake of breath, and overall audio quality. It is established that there is an improvement in several deepfake detection algorithms when they augmented the traditional and common features of audio data with these EDLFs. In this paper, using a hybrid dataset comprised of multiple types of spoofed audio augmented with sociolinguistic annotations, we investigate causal discovery and inferences between the discernible linguistic features and the label in the audio clips, comparing the findings of the causal models with the expert ground truth validation labeling process. Our findings suggest that the causal models indicate the utility of incorporating linguistic features to help discern spoofed audio, as well as the overall need and opportunity to incorporate human knowledge into models and techniques for strengthening AI models. The causal discovery and inference can be used as a foundation of training humans to discern spoofed audio as well as automating EDLFs labeling for the purpose of performance improvement of the common AI-based spoofed audio detectors.
Title:
DetoxBench: Benchmarking Large Language Models for Multitask Fraud & Abuse Detection
Abstract
Large language models (LLMs) have demonstrated remarkable capabilities in natural language processing tasks. However, their practical application in high-stake domains, such as fraud and abuse detection, remains an area that requires further exploration. The existing applications often narrowly focus on specific tasks like toxicity or hate speech detection. In this paper, we present a comprehensive benchmark suite designed to assess the performance of LLMs in identifying and mitigating fraudulent and abusive language across various real-world scenarios. Our benchmark encompasses a diverse set of tasks, including detecting spam emails, hate speech, misogynistic language, and more. We evaluated several state-of-the-art LLMs, including models from Anthropic, Mistral AI, and the AI21 family, to provide a comprehensive assessment of their capabilities in this critical domain. The results indicate that while LLMs exhibit proficient baseline performance in individual fraud and abuse detection tasks, their performance varies considerably across tasks, particularly struggling with tasks that demand nuanced pragmatic reasoning, such as identifying diverse forms of misogynistic language. These findings have important implications for the responsible development and deployment of LLMs in high-risk applications. Our benchmark suite can serve as a tool for researchers and practitioners to systematically evaluate LLMs for multi-task fraud detection and drive the creation of more robust, trustworthy, and ethically-aligned systems for fraud and abuse detection.
Title:
Symmetry constrained neural networks for detection and localization of damage in metal plates
Authors: James Amarel, Christopher Rudolf, Athanasios Iliopoulos, John Michopoulos, Leslie N. Smith
Subjects: Subjects:
Machine Learning (cs.LG); Signal Processing (eess.SP)
Abstract
The present paper is concerned with deep learning techniques applied to detection and localization of damage in a thin aluminum plate. We used data generated on a tabletop apparatus by mounting to the plate four piezoelectric transducers, each of which took turn to generate a Lamb wave that then traversed the region of interest before being received by the remaining three sensors. On training a neural network to analyze time-series data of the material response, which displayed damage-reflective features whenever the plate guided waves interacted with a contact load, we achieved a model that detected with greater than 99% accuracy in addition to a model that localized with $3.14 \pm 0.21$ mm mean distance error and captured more than 60% of test examples within the diffraction limit. For each task, the best-performing model was designed according to the inductive bias that our transducers were both similar and arranged in a square pattern on a nearly uniform plate.
Title:
Loss Distillation via Gradient Matching for Point Cloud Completion with Weighted Chamfer Distance
Authors: Fangzhou Lin, Haotian Liu, Haoying Zhou, Songlin Hou, Kazunori D Yamada, Gregory S. Fischer, Yanhua Li, Haichong K. Zhang, Ziming Zhang
Abstract
3D point clouds enhanced the robot's ability to perceive the geometrical information of the environments, making it possible for many downstream tasks such as grasp pose detection and scene understanding. The performance of these tasks, though, heavily relies on the quality of data input, as incomplete can lead to poor results and failure cases. Recent training loss functions designed for deep learning-based point cloud completion, such as Chamfer distance (CD) and its variants (\eg HyperCD ), imply a good gradient weighting scheme can significantly boost performance. However, these CD-based loss functions usually require data-related parameter tuning, which can be time-consuming for data-extensive tasks. To address this issue, we aim to find a family of weighted training losses ({\em weighted CD}) that requires no parameter tuning. To this end, we propose a search scheme, {\em Loss Distillation via Gradient Matching}, to find good candidate loss functions by mimicking the learning behavior in backpropagation between HyperCD and weighted CD. Once this is done, we propose a novel bilevel optimization formula to train the backbone network based on the weighted CD loss. We observe that: (1) with proper weighted functions, the weighted CD can always achieve similar performance to HyperCD, and (2) the Landau weighted CD, namely {\em Landau CD}, can outperform HyperCD for point cloud completion and lead to new state-of-the-art results on several benchmark datasets. {\it Our demo code is available at \url{this https URL}.}
Title:
Development and Benchmarking of Multilingual Code Clone Detector
Abstract
The diversity of programming languages is growing, making the language extensibility of code clone detectors crucial. However, this is challenging for most existing clone detection detectors because the source code handler needs modifications, which require specialist-level knowledge of the targeted language and is time-consuming. Multilingual code clone detectors make it easier to add new language support by providing syntax information of the target language only. To address the shortcomings of existing multilingual detectors for language scalability and detection performance, we propose a multilingual code block extraction method based on ANTLR parser generation, and implement a multilingual code clone detector (MSCCD), which supports the most significant number of languages currently available and has the ability to detect Type-3 code clones. We follow the methodology of previous studies to evaluate the detection performance of the Java language. Compared to ten state-of-the-art detectors, MSCCD performs at an average level while it also supports a significantly larger number of languages. Furthermore, we propose the first multilingual syntactic code clone evaluation benchmark based on the CodeNet database. Our results reveal that even when applying the same detection approach, performance can vary markedly depending on the language of the source code under investigation. Overall, MSCCD is the most balanced one among the evaluated tools when considering detection performance and language extensibility.
Title:
MTDA-HSED: Mutual-Assistance Tuning and Dual-Branch Aggregating for Heterogeneous Sound Event Detection
Authors: Zehao Wang, Haobo Yue, Zhicheng Zhang, Da Mu, Jin Tang, Jianqin Yin
Abstract
Sound Event Detection (SED) plays a vital role in comprehending and perceiving acoustic scenes. Previous methods have demonstrated impressive capabilities. However, they are deficient in learning features of complex scenes from heterogeneous dataset. In this paper, we introduce a novel dual-branch architecture named Mutual-Assistance Tuning and Dual-Branch Aggregating for Heterogeneous Sound Event Detection (MTDA-HSED). The MTDA-HSED architecture employs the Mutual-Assistance Audio Adapter (M3A) to effectively tackle the multi-scenario problem and uses the Dual-Branch Mid-Fusion (DBMF) module to tackle the multi-granularity problem. Specifically, M3A is integrated into the BEATs block as an adapter to improve the BEATs' performance by fine-tuning it on the multi-scenario dataset. The DBMF module connects BEATs and CNN branches, which facilitates the deep fusion of information from the BEATs and the CNN branches. Experimental results show that the proposed methods exceed the baseline of mpAUC by \textbf{$5\%$} on the DESED and MAESTRO Real datasets. Code is \href{this https URL}{here}.
Abstract
Scene Change Detection (SCD) is vital for applications such as visual surveillance and mobile robotics. However, current SCD methods exhibit a bias to the temporal order of training datasets and limited performance on unseen domains; coventional SCD benchmarks are not able to evaluate generalization or temporal consistency. To tackle these limitations, we introduce a Generalizable Scene Change Detection Framework (GeSCF) in this work. The proposed GeSCF leverages localized semantics of a foundation model without any re-training or fine-tuning -- for generalization over unseen domains. Specifically, we design an adaptive thresholding of the similarity distribution derived from facets of the pre-trained foundation model to generate initial pseudo-change mask. We further utilize Segment Anything Model's (SAM) class-agnostic masks to refine pseudo-masks. Moreover, our proposed framework maintains commutative operations in all settings to ensure complete temporal consistency. Finally, we define new metrics, evaluation dataset, and evaluation protocol for Generalizable Scene Change Detection (GeSCD). Extensive experiments demonstrate that GeSCF excels across diverse and challenging environments -- establishing a new benchmark for SCD performance.
Title:
SubRegWeigh: Effective and Efficient Annotation Weighing with Subword Regularization
Abstract
Many datasets of natural language processing (NLP) sometimes include annotation errors. Researchers have attempted to develop methods to reduce the adverse effect of errors in datasets automatically. However, an existing method is time-consuming because it requires many trained models to detect errors. We propose a novel method to reduce the time of error detection. Specifically, we use a tokenization technique called subword regularization to create pseudo-multiple models which are used to detect errors. Our proposed method, SubRegWeigh, can perform annotation weighting four to five times faster than the existing method. Additionally, SubRegWeigh improved performance in both document classification and named entity recognition tasks. In experiments with pseudo-incorrect labels, pseudo-incorrect labels were adequately detected.
Title:
ALSS-YOLO: An Adaptive Lightweight Channel Split and Shuffling Network for TIR Wildlife Detection in UAV Imagery
Authors: Ang He, Xiaobo Li, Ximei Wu, Chengyue Su, Jing Chen, Sheng Xu, Xiaobin Guo
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Unmanned aerial vehicles (UAVs) equipped with thermal infrared (TIR) cameras play a crucial role in combating nocturnal wildlife poaching. However, TIR images often face challenges such as jitter, and wildlife overlap, necessitating UAVs to possess the capability to identify blurred and overlapping small targets. Current traditional lightweight networks deployed on UAVs struggle to extract features from blurry small targets. To address this issue, we developed ALSS-YOLO, an efficient and lightweight detector optimized for TIR aerial images. Firstly, we propose a novel Adaptive Lightweight Channel Split and Shuffling (ALSS) module. This module employs an adaptive channel split strategy to optimize feature extraction and integrates a channel shuffling mechanism to enhance information exchange between channels. This improves the extraction of blurry features, crucial for handling jitter-induced blur and overlapping targets. Secondly, we developed a Lightweight Coordinate Attention (LCA) module that employs adaptive pooling and grouped convolution to integrate feature information across dimensions. This module ensures lightweight operation while maintaining high detection precision and robustness against jitter and target overlap. Additionally, we developed a single-channel focus module to aggregate the width and height information of each channel into four-dimensional channel fusion, which improves the feature representation efficiency of infrared images. Finally, we modify the localization loss function to emphasize the loss value associated with small objects to improve localization accuracy. Extensive experiments on the BIRDSAI and ISOD TIR UAV wildlife datasets show that ALSS-YOLO achieves state-of-the-art performance, Our code is openly available at this https URL.
Title:
Context Enhancement with Reconstruction as Sequence for Unified Unsupervised Anomaly Detection
Abstract
Unsupervised anomaly detection (AD) aims to train robust detection models using only normal samples, while can generalize well to unseen anomalies. Recent research focuses on a unified unsupervised AD setting in which only one model is trained for all classes, i.e., n-class-one-model paradigm. Feature-reconstruction-based methods achieve state-of-the-art performance in this scenario. However, existing methods often suffer from a lack of sufficient contextual awareness, thereby compromising the quality of the reconstruction. To address this issue, we introduce a novel Reconstruction as Sequence (RAS) method, which enhances the contextual correspondence during feature reconstruction from a sequence modeling perspective. In particular, based on the transformer technique, we integrate a specialized RASFormer block into RAS. This block enables the capture of spatial relationships among different image regions and enhances sequential dependencies throughout the reconstruction process. By incorporating the RASFormer block, our RAS method achieves superior contextual awareness capabilities, leading to remarkable performance. Experimental results show that our RAS significantly outperforms competing methods, well demonstrating the effectiveness and superiority of our method. Our code is available at this https URL.
Title:
An Attribute-Enriched Dataset and Auto-Annotated Pipeline for Open Detection
Abstract
Detecting objects of interest through language often presents challenges, particularly with objects that are uncommon or complex to describe, due to perceptual discrepancies between automated models and human annotators. These challenges highlight the need for comprehensive datasets that go beyond standard object labels by incorporating detailed attribute descriptions. To address this need, we introduce the Objects365-Attr dataset, an extension of the existing Objects365 dataset, distinguished by its attribute annotations. This dataset reduces inconsistencies in object detection by integrating a broad spectrum of attributes, including color, material, state, texture and tone. It contains an extensive collection of 5.6M object-level attribute descriptions, meticulously annotated across 1.4M bounding boxes. Additionally, to validate the dataset's effectiveness, we conduct a rigorous evaluation of YOLO-World at different scales, measuring their detection performance and demonstrating the dataset's contribution to advancing object detection.
Title:
SDF-Net: A Hybrid Detection Network for Mediastinal Lymph Node Detection on Contrast CT Images
Abstract
Accurate lymph node detection and quantification are crucial for cancer diagnosis and staging on contrast-enhanced CT images, as they impact treatment planning and prognosis. However, detecting lymph nodes in the mediastinal area poses challenges due to their low contrast, irregular shapes and dispersed distribution. In this paper, we propose a Swin-Det Fusion Network (SDF-Net) to effectively detect lymph nodes. SDF-Net integrates features from both segmentation and detection to enhance the detection capability of lymph nodes with various shapes and sizes. Specifically, an auto-fusion module is designed to merge the feature maps of segmentation and detection networks at different levels. To facilitate effective learning without mask annotations, we introduce a shape-adaptive Gaussian kernel to represent lymph node in the training stage and provide more anatomical information for effective learning. Comparative results demonstrate promising performance in addressing the complex lymph node detection problem.
Abstract
With the rapid advancement of technologies like text-to-speech (TTS) and voice conversion (VC), detecting deepfake voices has become increasingly crucial. However, both academia and industry lack a comprehensive and intuitive benchmark for evaluating detectors. Existing datasets are limited in language diversity and lack many manipulations encountered in real-world production environments. To fill this gap, we propose VoiceWukong, a benchmark designed to evaluate the performance of deepfake voice detectors. To build the dataset, we first collected deepfake voices generated by 19 advanced and widely recognized commercial tools and 15 open-source tools. We then created 38 data variants covering six types of manipulations, constructing the evaluation dataset for deepfake voice detection. VoiceWukong thus includes 265,200 English and 148,200 Chinese deepfake voice samples. Using VoiceWukong, we evaluated 12 state-of-the-art detectors. AASIST2 achieved the best equal error rate (EER) of 13.50%, while all others exceeded 20%. Our findings reveal that these detectors face significant challenges in real-world applications, with dramatically declining performance. In addition, we conducted a user study with more than 300 participants. The results are compared with the performance of the 12 detectors and a multimodel large language model (MLLM), i.e., Qwen2-Audio, where different detectors and humans exhibit varying identification capabilities for deepfake voices at different deception levels, while the LALM demonstrates no detection ability at all. Furthermore, we provide a leaderboard for deepfake voice detection, publicly available at {this https URL}.
Title:
Texture-AD: An Anomaly Detection Dataset and Benchmark for Real Algorithm Development
Abstract
Anomaly detection is a crucial process in industrial manufacturing and has made significant advancements recently. However, there is a large variance between the data used in the development and the data collected by the production environment. Therefore, we present the Texture-AD benchmark based on representative texture-based anomaly detection to evaluate the effectiveness of unsupervised anomaly detection algorithms in real-world applications. This dataset includes images of 15 different cloth, 14 semiconductor wafers and 10 metal plates acquired under different optical schemes. In addition, it includes more than 10 different types of defects produced during real manufacturing processes, such as scratches, wrinkles, color variations and point defects, which are often more difficult to detect than existing datasets. All anomalous areas are provided with pixel-level annotations to facilitate comprehensive evaluation using anomaly detection models. Specifically, to adapt to diverse products in automated pipelines, we present a new evaluation method and results of baseline algorithms. The experimental results show that Texture-AD is a difficult challenge for state-of-the-art algorithms. To our knowledge, Texture-AD is the first dataset to be devoted to evaluating industrial defect detection algorithms in the real world. The dataset is available at https://XXX.
Title:
The Impact of SBOM Generators on Vulnerability Assessment in Python: A Comparison and a Novel Approach
Abstract
The Software Supply Chain (SSC) security is a critical concern for both users and developers. Recent incidents, like the SolarWinds Orion compromise, proved the widespread impact resulting from the distribution of compromised software. The reliance on open-source components, which constitute a significant portion of modern software, further exacerbates this risk. To enhance SSC security, the Software Bill of Materials (SBOM) has been promoted as a tool to increase transparency and verifiability in software composition. However, despite its promise, SBOMs are not without limitations. Current SBOM generation tools often suffer from inaccuracies in identifying components and dependencies, leading to the creation of erroneous or incomplete representations of the SSC. Despite existing studies exposing these limitations, their impact on the vulnerability detection capabilities of security tools is still unknown. In this paper, we perform the first security analysis on the vulnerability detection capabilities of tools receiving SBOMs as input. We comprehensively evaluate SBOM generation tools by providing their outputs to vulnerability identification software. Based on our results, we identify the root causes of these tools' ineffectiveness and propose PIP-sbom, a novel pip-inspired solution that addresses their shortcomings. PIP-sbom provides improved accuracy in component identification and dependency resolution. Compared to best-performing state-of-the-art tools, PIP-sbom increases the average precision and recall by 60%, and reduces by ten times the number of false positives.
Title:
Soft Acoustic Curvature Sensor: Design and Development
Authors: Mohammad Sheikh Sofla, Hanita Golshanian, Vishnu Rajendran S, Amir Ghalamzan E
Abstract
This paper introduces a novel Soft Acoustic Curvature (SAC) sensor. SAC incorporates integrated audio components and features an acoustic channel within a flexible structure. A reference acoustic wave, generated by a speaker at one end of the channel, propagates and is received by a microphone at the other channel's end. Our previous study revealed that acoustic wave energy dissipation varies with acoustic channel deformation, leading us to design a novel channel capable of large deformation due to bending. We then use Machine Learning (ML) models to establish a complex mapping between channel deformations and sound modulation. Various sound frequencies and ML models were evaluated to enhance curvature detection accuracy. The sensor, constructed using soft material and 3D printing, was validated experimentally, with curvature measurement errors remaining within 3.5 m-1 for a range of 0 to 60 m-1 curvatures. These results demonstrate the effectiveness of the proposed method for estimating curvatures. With its flexible structure, the SAC sensor holds potential for applications in soft robotics, including shape measurement for continuum manipulators, soft grippers, and wearable devices.
Title:
A Likelihood Ratio-Based Approach to Segmenting Unknown Objects
Authors: Nazir Nayal, Youssef Shoeb, Fatma Güney
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Addressing the Out-of-Distribution (OoD) segmentation task is a prerequisite for perception systems operating in an open-world environment. Large foundational models are frequently used in downstream tasks, however, their potential for OoD remains mostly unexplored. We seek to leverage a large foundational model to achieve robust representation. Outlier supervision is a widely used strategy for improving OoD detection of the existing segmentation networks. However, current approaches for outlier supervision involve retraining parts of the original network, which is typically disruptive to the model's learned feature representation. Furthermore, retraining becomes infeasible in the case of large foundational models. Our goal is to retrain for outlier segmentation without compromising the strong representation space of the foundational model. To this end, we propose an adaptive, lightweight unknown estimation module (UEM) for outlier supervision that significantly enhances the OoD segmentation performance without affecting the learned feature representation of the original network. UEM learns a distribution for outliers and a generic distribution for known classes. Using the learned distributions, we propose a likelihood-ratio-based outlier scoring function that fuses the confidence of UEM with that of the pixel-wise segmentation inlier network to detect unknown objects. We also propose an objective to optimize this score directly. Our approach achieves a new state-of-the-art across multiple datasets, outperforming the previous best method by 5.74% average precision points while having a lower false-positive rate. Importantly, strong inlier performance remains unaffected.
Title:
GeMuCo: Generalized Multisensory Correlational Model for Body Schema Learning
Abstract
Humans can autonomously learn the relationship between sensation and motion in their own bodies, estimate and control their own body states, and move while continuously adapting to the current environment. On the other hand, current robots control their bodies by learning the network structure described by humans from their experiences, making certain assumptions on the relationship between sensors and actuators. In addition, the network model does not adapt to changes in the robot's body, the tools that are grasped, or the environment, and there is no unified theory, not only for control but also for state estimation, anomaly detection, simulation, and so on. In this study, we propose a Generalized Multisensory Correlational Model (GeMuCo), in which the robot itself acquires a body schema describing the correlation between sensors and actuators from its own experience, including model structures such as network input/output. The robot adapts to the current environment by updating this body schema model online, estimates and controls its body state, and even performs anomaly detection and simulation. We demonstrate the effectiveness of this method by applying it to tool-use considering changes in grasping state for an axis-driven robot, to joint-muscle mapping learning for a musculoskeletal robot, and to full-body tool manipulation for a low-rigidity plastic-made humanoid.
Title:
Human-mimetic binaural ear design and sound source direction estimation for task realization of musculoskeletal humanoids
Abstract
Human-like environment recognition by musculoskeletal humanoids is important for task realization in real complex environments and for use as dummies for test subjects. Humans integrate various sensory information to perceive their surroundings, and hearing is particularly useful for recognizing objects out of view or out of touch. In this research, we aim to realize human-like auditory environmental recognition and task realization for musculoskeletal humanoids by equipping them with a human-like auditory processing system. Humans realize sound-based environmental recognition by estimating directions of the sound sources and detecting environmental sounds based on changes in the time and frequency domain of incoming sounds and the integration of auditory information in the central nervous system. We propose a human mimetic auditory information processing system, which consists of three components: the human mimetic binaural ear unit, which mimics human ear structure and characteristics, the sound source direction estimation system, and the environmental sound detection system, which mimics processing in the central nervous system. We apply it to Musashi, a human mimetic musculoskeletal humanoid, and have it perform tasks that require sound information outside of view in real noisy environments to confirm the usefulness of the proposed methods.
Title:
Knowledge Distillation via Query Selection for Detection Transformer
Authors: Yi Liu, Luting Wang, Zongheng Tang, Yue Liao, Yifan Sun, Lijun Zhang, Si Liu
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Transformers have revolutionized the object detection landscape by introducing DETRs, acclaimed for their simplicity and efficacy. Despite their advantages, the substantial size of these models poses significant challenges for practical deployment, particularly in resource-constrained environments. This paper addresses the challenge of compressing DETR by leveraging knowledge distillation, a technique that holds promise for maintaining model performance while reducing size. A critical aspect of DETRs' performance is their reliance on queries to interpret object representations accurately. Traditional distillation methods often focus exclusively on positive queries, identified through bipartite matching, neglecting the rich information present in hard-negative queries. Our visual analysis indicates that hard-negative queries, focusing on foreground elements, are crucial for enhancing distillation outcomes. To this end, we introduce a novel Group Query Selection strategy, which diverges from traditional query selection in DETR distillation by segmenting queries based on their Generalized Intersection over Union (GIoU) with ground truth objects, thereby uncovering valuable hard-negative queries for distillation. Furthermore, we present the Knowledge Distillation via Query Selection for DETR (QSKD) framework, which incorporates Attention-Guided Feature Distillation (AGFD) and Local Alignment Prediction Distillation (LAPD). These components optimize the distillation process by focusing on the most informative aspects of the teacher model's intermediate features and output. Our comprehensive experimental evaluation of the MS-COCO dataset demonstrates the effectiveness of our approach, significantly improving average precision (AP) across various DETR architectures without incurring substantial computational costs. Specifically, the AP of Conditional DETR ResNet-18 increased from 35.8 to 39.9.
Title:
A Machine Learning Based Approach for Statistical Analysis of Detonation Cells from Soot Foils
Authors: Vansh Sharma, Michael Ullman, Venkat Raman
Abstract
This study presents a novel algorithm based on machine learning (ML) for the precise segmentation and measurement of detonation cells from soot foil images, addressing the limitations of manual and primitive edge detection methods prevalent in the field. Using advances in cellular biology segmentation models, the proposed algorithm is designed to accurately extract cellular patterns without a training procedure or dataset, which is a significant challenge in detonation research. The algorithm's performance was validated using a series of test cases that mimic experimental and numerical detonation studies. The results demonstrated consistent accuracy, with errors remaining within 10%, even in complex cases. The algorithm effectively captured key cell metrics such as cell area and span, revealing trends across different soot foil samples with uniform to highly irregular cellular structures. Although the model proved robust, challenges remain in segmenting and analyzing highly complex or irregular cellular patterns. This work highlights the broad applicability and potential of the algorithm to advance the understanding of detonation wave dynamics.
Title:
UAVDB: Trajectory-Guided Adaptable Bounding Boxes for UAV Detection
Authors: Yu-Hsi Chen
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV); Applications (stat.AP)
Abstract
With the rapid development of drone technology, accurate detection of Unmanned Aerial Vehicles (UAVs) has become essential for applications such as surveillance, security, and airspace management. In this paper, we propose a novel trajectory-guided method, the Patch Intensity Convergence (PIC) technique, which generates high-fidelity bounding boxes for UAV detection tasks and no need for the effort required for labeling. The PIC technique forms the foundation for developing UAVDB, a database explicitly created for UAV detection. Unlike existing datasets, which often use low-resolution footage or focus on UAVs in simple backgrounds, UAVDB employs high-resolution video to capture UAVs at various scales, ranging from hundreds of pixels to nearly single-digit sizes. This broad-scale variation enables comprehensive evaluation of detection algorithms across different UAV sizes and distances. Applying the PIC technique, we can also efficiently generate detection datasets from trajectory or positional data, even without size information. We extensively benchmark UAVDB using YOLOv8 series detectors, offering a detailed performance analysis. Our findings highlight UAVDB's potential as a vital database for advancing UAV detection, particularly in high-resolution and long-distance tracking scenarios.
Title:
Semi-Supervised 3D Object Detection with Chanel Augmentation using Transformation Equivariance
Authors: Minju Kang, Taehun Kong, Tae-Kyun Kim
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Accurate 3D object detection is crucial for autonomous vehicles and robots to navigate and interact with the environment safely and effectively. Meanwhile, the performance of 3D detector relies on the data size and annotation which is expensive. Consequently, the demand of training with limited labeled data is growing. We explore a novel teacher-student framework employing channel augmentation for 3D semi-supervised object detection. The teacher-student SSL typically adopts a weak augmentation and strong augmentation to teacher and student, respectively. In this work, we apply multiple channel augmentations to both networks using the transformation equivariance detector (TED). The TED allows us to explore different combinations of augmentation on point clouds and efficiently aggregates multi-channel transformation equivariance features. In principle, by adopting fixed channel augmentations for the teacher network, the student can train stably on reliable pseudo-labels. Adopting strong channel augmentations can enrich the diversity of data, fostering robustness to transformations and enhancing generalization performance of the student network. We use SOTA hierarchical supervision as a baseline and adapt its dual-threshold to TED, which is called channel IoU consistency. We evaluate our method with KITTI dataset, and achieved a significant performance leap, surpassing SOTA 3D semi-supervised object detection models.
Title:
Transtreaming: Adaptive Delay-aware Transformer for Real-time Streaming Perception
Abstract
Real-time object detection is critical for the decision-making process for many real-world applications, such as collision avoidance and path planning in autonomous driving. This work presents an innovative real-time streaming perception method, Transtreaming, which addresses the challenge of real-time object detection with dynamic computational delay. The core innovation of Transtreaming lies in its adaptive delay-aware transformer, which can concurrently predict multiple future frames and select the output that best matches the real-world present time, compensating for any system-induced computation delays. The proposed model outperforms the existing state-of-the-art methods, even in single-frame detection scenarios, by leveraging a transformer-based methodology. It demonstrates robust performance across a range of devices, from powerful V100 to modest 2080Ti, achieving the highest level of perceptual accuracy on all platforms. Unlike most state-of-the-art methods that struggle to complete computation within a single frame on less powerful devices, Transtreaming meets the stringent real-time processing requirements on all kinds of devices. The experimental results emphasize the system's adaptability and its potential to significantly improve the safety and reliability for many real-world systems, such as autonomous driving.
Title:
GroUSE: A Benchmark to Evaluate Evaluators in Grounded Question Answering
Abstract
Retrieval-Augmented Generation (RAG) has emerged as a common paradigm to use Large Language Models (LLMs) alongside private and up-to-date knowledge bases. In this work, we address the challenges of using LLM-as-a-Judge when evaluating grounded answers generated by RAG systems. To assess the calibration and discrimination capabilities of judge models, we identify 7 generator failure modes and introduce GroUSE (Grounded QA Unitary Scoring of Evaluators), a meta-evaluation benchmark of 144 unit tests. This benchmark reveals that existing automated RAG evaluation frameworks often overlook important failure modes, even when using GPT-4 as a judge. To improve on the current design of automated RAG evaluation frameworks, we propose a novel pipeline and find that while closed models perform well on GroUSE, state-of-the-art open-source judges do not generalize to our proposed criteria, despite strong correlation with GPT-4's judgement. Our findings suggest that correlation with GPT-4 is an incomplete proxy for the practical performance of judge models and should be supplemented with evaluations on unit tests for precise failure mode detection. We further show that finetuning Llama-3 on GPT-4's reasoning traces significantly boosts its evaluation capabilities, improving upon both correlation with GPT-4's evaluations and calibration on reference situations.
Title:
When to Extract ReID Features: A Selective Approach for Improved Multiple Object Tracking
Authors: Emirhan Bayar, Cemal Aker
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Extracting and matching Re-Identification (ReID) features is used by many state-of-the-art (SOTA) Multiple Object Tracking (MOT) methods, particularly effective against frequent and long-term occlusions. While end-to-end object detection and tracking have been the main focus of recent research, they have yet to outperform traditional methods in benchmarks like MOT17 and MOT20. Thus, from an application standpoint, methods with separate detection and embedding remain the best option for accuracy, modularity, and ease of implementation, though they are impractical for edge devices due to the overhead involved. In this paper, we investigate a selective approach to minimize the overhead of feature extraction while preserving accuracy, modularity, and ease of implementation. This approach can be integrated into various SOTA methods. We demonstrate its effectiveness by applying it to StrongSORT and Deep OC-SORT. Experiments on MOT17, MOT20, and DanceTrack datasets show that our mechanism retains the advantages of feature extraction during occlusions while significantly reducing runtime. Additionally, it improves accuracy by preventing confusion in the feature-matching stage, particularly in cases of deformation and appearance similarity, which are common in DanceTrack. this https URL, this https URL
Title:
Towards Localizing Structural Elements: Merging Geometrical Detection with Semantic Verification in RGB-D Data
Authors: Ali Tourani, Saad Ejaz, Hriday Bavle, Jose Luis Sanchez-Lopez, Holger Voos
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
Abstract
RGB-D cameras supply rich and dense visual and spatial information for various robotics tasks such as scene understanding, map reconstruction, and localization. Integrating depth and visual information can aid robots in localization and element mapping, advancing applications like 3D scene graph generation and Visual Simultaneous Localization and Mapping (VSLAM). While point cloud data containing such information is primarily used for enhanced scene understanding, exploiting their potential to capture and represent rich semantic information has yet to be adequately targeted. This paper presents a real-time pipeline for localizing building components, including wall and ground surfaces, by integrating geometric calculations for pure 3D plane detection followed by validating their semantic category using point cloud data from RGB-D cameras. It has a parallel multi-thread architecture to precisely estimate poses and equations of all the planes detected in the environment, filters the ones forming the map structure using a panoptic segmentation validation, and keeps only the validated building components. Incorporating the proposed method into a VSLAM framework confirmed that constraining the map with the detected environment-driven semantic elements can improve scene understanding and map reconstruction accuracy. It can also ensure (re-)association of these detected components into a unified 3D scene graph, bridging the gap between geometric accuracy and semantic understanding. Additionally, the pipeline allows for the detection of potential higher-level structural entities, such as rooms, by identifying the relationships between building components based on their layout.
Title:
EyeCLIP: A visual-language foundation model for multi-modal ophthalmic image analysis
Abstract
Early detection of eye diseases like glaucoma, macular degeneration, and diabetic retinopathy is crucial for preventing vision loss. While artificial intelligence (AI) foundation models hold significant promise for addressing these challenges, existing ophthalmic foundation models primarily focus on a single modality, whereas diagnosing eye diseases requires multiple modalities. A critical yet often overlooked aspect is harnessing the multi-view information across various modalities for the same patient. Additionally, due to the long-tail nature of ophthalmic diseases, standard fully supervised or unsupervised learning approaches often struggle. Therefore, it is essential to integrate clinical text to capture a broader spectrum of diseases. We propose EyeCLIP, a visual-language foundation model developed using over 2.77 million multi-modal ophthalmology images with partial text data. To fully leverage the large multi-modal unlabeled and labeled data, we introduced a pretraining strategy that combines self-supervised reconstructions, multi-modal image contrastive learning, and image-text contrastive learning to learn a shared representation of multiple modalities. Through evaluation using 14 benchmark datasets, EyeCLIP can be transferred to a wide range of downstream tasks involving ocular and systemic diseases, achieving state-of-the-art performance in disease classification, visual question answering, and cross-modal retrieval. EyeCLIP represents a significant advancement over previous methods, especially showcasing few-shot, even zero-shot capabilities in real-world long-tail scenarios.
Title:
A Semantic Segmentation Approach on Sweet Orange Leaf Diseases Detection Utilizing YOLO
Authors: Sabit Ahamed Preanto (4IR Research Cell Daffodil International University, Dhaka, Bangladesh), Md. Taimur Ahad (4IR Research Cell Daffodil International University, Dhaka, Bangladesh), Yousuf Rayhan Emon (4IR Research Cell Daffodil International University, Dhaka, Bangladesh), Sumaya Mustofa (4IR Research Cell Daffodil International University, Dhaka, Bangladesh), Md Alamin (4IR Research Cell Daffodil International University, Dhaka, Bangladesh)
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
This research introduces an advanced method for diagnosing diseases in sweet orange leaves by utilising advanced artificial intelligence models like YOLOv8 . Due to their significance as a vital agricultural product, sweet oranges encounter significant threats from a variety of diseases that harmfully affect both their yield and quality. Conventional methods for disease detection primarily depend on manual inspection which is ineffective and frequently leads to errors, resulting in delayed treatment and increased financial losses. In response to this challenge, the research utilized YOLOv8 , harnessing their proficiencies in detecting objects and analyzing images. YOLOv8 is recognized for its rapid and precise performance, while VIT is acknowledged for its detailed feature extraction abilities. Impressively, during both the training and validation stages, YOLOv8 exhibited a perfect accuracy of 80.4%, while VIT achieved an accuracy of 99.12%, showcasing their potential to transform disease detection in agriculture. The study comprehensively examined the practical challenges related to the implementation of AI technologies in agriculture, encompassing the computational demands and user accessibility, and offering viable solutions for broader usage. Moreover, it underscores the environmental considerations, particularly the potential for reduced pesticide usage, thereby promoting sustainable farming and environmental conservation. These findings provide encouraging insights into the application of AI in agriculture, suggesting a transition towards more effective, sustainable, and technologically advanced farming methods. This research not only highlights the efficacy of YOLOv8 within a specific agricultural domain but also lays the foundation for further studies that encompass a broader application in crop management and sustainable agricultural practices.
Title:
Technical Report of Mobile Manipulator Robot for Industrial Environments
Authors: Erfan Amoozad Khalili, Kiarash Ghasemzadeh, Hossein Gohari, Mohammadreza Jafari, Matin Jamshidi, Mahdi Khaksar, AmirReza AkramiFard, Mana Hatamzadeh, Saba Sadeghi, Mohammad Hossein Moaiyeri
Abstract
This paper presents the development of the Auriga @Work robot, designed by the Robotics and Intelligent Automation Lab at Shahid Beheshti University, Department of Electrical Engineering, for the RoboCup 2024 competition. The robot is tailored for industrial applications, focusing on enhancing efficiency in repetitive or hazardous environments. It is equipped with a 4-wheel Mecanum drive system for omnidirectional mobility and a 5-degree-of-freedom manipulator arm with a custom 3D-printed gripper for object manipulation and navigation tasks. The robot's electronics are powered by custom-designed boards utilizing ESP32 microcontrollers and an Nvidia Jetson Nano for real-time control and decision-making. The key software stack integrates Hector SLAM for mapping, the A* algorithm for path planning, and YOLO for object detection, along with advanced sensor fusion for improved navigation and collision avoidance.
Keyword: face recognition
Title:
Distilling Generative-Discriminative Representations for Very Low-Resolution Face Recognition
Abstract
Very low-resolution face recognition is challenging due to the serious loss of informative facial details in resolution degradation. In this paper, we propose a generative-discriminative representation distillation approach that combines generative representation with cross-resolution aligned knowledge distillation. This approach facilitates very low-resolution face recognition by jointly distilling generative and discriminative models via two distillation modules. Firstly, the generative representation distillation takes the encoder of a diffusion model pretrained for face super-resolution as the generative teacher to supervise the learning of the student backbone via feature regression, and then freezes the student backbone. After that, the discriminative representation distillation further considers a pretrained face recognizer as the discriminative teacher to supervise the learning of the student head via cross-resolution relational contrastive distillation. In this way, the general backbone representation can be transformed into discriminative head representation, leading to a robust and discriminative student model for very low-resolution face recognition. Our approach improves the recovery of the missing details in very low-resolution faces and achieves better knowledge transfer. Extensive experiments on face datasets demonstrate that our approach enhances the recognition accuracy of very low-resolution faces, showcasing its effectiveness and adaptability.
Keyword: augmentation
Title:
Continuous Learning of Transformer-based Audio Deepfake Detection
Authors: Tuan Duy Nguyen Le, Kah Kuan Teh, Huy Dat Tran
Subjects: Subjects:
Sound (cs.SD); Audio and Speech Processing (eess.AS)
Abstract
This paper proposes a novel framework for audio deepfake detection with two main objectives: i) attaining the highest possible accuracy on available fake data, and ii) effectively performing continuous learning on new fake data in a few-shot learning manner. Specifically, we conduct a large audio deepfake collection using various deep audio generation methods. The data is further enhanced with additional augmentation methods to increase variations amidst compressions, far-field recordings, noise, and other distortions. We then adopt the Audio Spectrogram Transformer for the audio deepfake detection model. Accordingly, the proposed method achieves promising performance on various benchmark datasets. Furthermore, we present a continuous learning plugin module to update the trained model most effectively with the fewest possible labeled data points of the new fake type. The proposed method outperforms the conventional direct fine-tuning approach with much fewer labeled data points.
Title:
A Small Claims Court for the NLP: Judging Legal Text Classification Strategies With Small Datasets
Abstract
Recent advances in language modelling has significantly decreased the need of labelled data in text classification tasks. Transformer-based models, pre-trained on unlabeled data, can outmatch the performance of models trained from scratch for each task. However, the amount of labelled data need to fine-tune such type of model is still considerably high for domains requiring expert-level annotators, like the legal domain. This paper investigates the best strategies for optimizing the use of a small labeled dataset and large amounts of unlabeled data and perform a classification task in the legal area with 50 predefined topics. More specifically, we use the records of demands to a Brazilian Public Prosecutor's Office aiming to assign the descriptions in one of the subjects, which currently demands deep legal knowledge for manual filling. The task of optimizing the performance of classifiers in this scenario is especially challenging, given the low amount of resources available regarding the Portuguese language, especially in the legal domain. Our results demonstrate that classic supervised models such as logistic regression and SVM and the ensembles random forest and gradient boosting achieve better performance along with embeddings extracted with word2vec when compared to BERT language model. The latter demonstrates superior performance in association with the architecture of the model itself as a classifier, having surpassed all previous models in that regard. The best result was obtained with Unsupervised Data Augmentation (UDA), which jointly uses BERT, data augmentation, and strategies of semi-supervised learning, with an accuracy of 80.7% in the aforementioned task.
Title:
Enhanced Generative Data Augmentation for Semantic Segmentation via Stronger Guidance
Authors: Quang-Huy Che, Duc-Tri Le, Vinh-Tiep Nguyen
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Data augmentation is a widely used technique for creating training data for tasks that require labeled data, such as semantic segmentation. This method benefits pixel-wise annotation tasks requiring much effort and intensive labor. Traditional data augmentation methods involve simple transformations like rotations and flips to create new images from existing ones. However, these new images may lack diversity along the main semantic axes in the data and not change high-level semantic properties. To address this issue, generative models have emerged as an effective solution for augmenting data by generating synthetic images. Controllable generative models offer a way to augment data for semantic segmentation tasks using a prompt and visual reference from the original image. However, using these models directly presents challenges, such as creating an effective prompt and visual reference to generate a synthetic image that accurately reflects the content and structure of the original. In this work, we introduce an effective data augmentation method for semantic segmentation using the Controllable Diffusion Model. Our proposed method includes efficient prompt generation using Class-Prompt Appending and Visual Prior Combination to enhance attention to labeled classes in real images. These techniques allow us to generate images that accurately depict segmented classes in the real image. In addition, we employ the class balancing algorithm to ensure efficiency when merging the synthetic and original images to generate balanced data for the training dataset. We evaluated our method on the PASCAL VOC datasets and found it highly effective for synthesizing images in semantic segmentation.
Title:
EDADepth: Enhanced Data Augmentation for Monocular Depth Estimation
Abstract
Due to their text-to-image synthesis feature, diffusion models have recently seen a rise in visual perception tasks, such as depth estimation. The lack of good-quality datasets makes the extraction of a fine-grain semantic context challenging for the diffusion models. The semantic context with fewer details further worsens the process of creating effective text embeddings that will be used as input for diffusion models. In this paper, we propose a novel EDADepth, an enhanced data augmentation method to estimate monocular depth without using additional training data. We use Swin2SR, a super-resolution model, to enhance the quality of input images. We employ the BEiT pre-trained semantic segmentation model for better extraction of text embeddings. We introduce BLIP-2 tokenizer to generate tokens from these text embeddings. The novelty of our approach is the introduction of Swin2SR, the BEiT model, and the BLIP-2 tokenizer in the diffusion-based pipeline for the monocular depth estimation. Our model achieves state-of-the-art results (SOTA) on the {\delta}3 metric on NYUv2 and KITTI datasets. It also achieves results comparable to those of the SOTA models in the RMSE and REL metrics. Finally, we also show improvements in the visualization of the estimated depth compared to the SOTA diffusion-based monocular depth estimation models. Code: this https URL.
Title:
RealisDance: Equip controllable character animation with realistic hands
Authors: Jingkai Zhou, Benzhi Wang, Weihua Chen, Jingqi Bai, Dongyang Li, Aixi Zhang, Hao Xu, Mingyang Yang, Fan Wang
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Controllable character animation is an emerging task that generates character videos controlled by pose sequences from given character images. Although character consistency has made significant progress via reference UNet, another crucial factor, pose control, has not been well studied by existing methods yet, resulting in several issues: 1) The generation may fail when the input pose sequence is corrupted. 2) The hands generated using the DWPose sequence are blurry and unrealistic. 3) The generated video will be shaky if the pose sequence is not smooth enough. In this paper, we present RealisDance to handle all the above issues. RealisDance adaptively leverages three types of poses, avoiding failed generation caused by corrupted pose sequences. Among these pose types, HaMeR provides accurate 3D and depth information of hands, enabling RealisDance to generate realistic hands even for complex gestures. Besides using temporal attention in the main UNet, RealisDance also inserts temporal attention into the pose guidance network, smoothing the video from the pose condition aspect. Moreover, we introduce pose shuffle augmentation during training to further improve generation robustness and video smoothness. Qualitative experiments demonstrate the superiority of RealisDance over other existing methods, especially in hand quality.
Abstract
Affordance denotes the potential interactions inherent in objects. The perception of affordance can enable intelligent agents to navigate and interact with new environments efficiently. Weakly supervised affordance grounding teaches agents the concept of affordance without costly pixel-level annotations, but with exocentric images. Although recent advances in weakly supervised affordance grounding yielded promising results, there remain challenges including the requirement for paired exocentric and egocentric image dataset, and the complexity in grounding diverse affordances for a single object. To address them, we propose INTeraction Relationship-aware weakly supervised Affordance grounding (INTRA). Unlike prior arts, INTRA recasts this problem as representation learning to identify unique features of interactions through contrastive learning with exocentric images only, eliminating the need for paired datasets. Moreover, we leverage vision-language model embeddings for performing affordance grounding flexibly with any text, designing text-conditioned affordance map generation to reflect interaction relationship for contrastive learning and enhancing robustness with our text synonym augmentation. Our method outperformed prior arts on diverse datasets such as AGD20K, IIT-AFF, CAD and UMD. Additionally, experimental results demonstrate that our method has remarkable domain scalability for synthesized images / illustrations and is capable of performing affordance grounding for novel interactions and objects.
Title:
Enhancing Temporal Understanding in Audio Question Answering for Large Audio Language Models
Authors: Arvind Krishna Sridhar, Yinyi Guo, Erik Visser
Subjects: Subjects:
Sound (cs.SD); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
Abstract
The Audio Question Answering task includes audio event classification, audio captioning, and open ended reasoning. Recently, Audio Question Answering has garnered attention due to the advent of Large Audio Language Models. Current literature focuses on constructing LALMs by integrating audio encoders with text only Large Language Models through a projection module. While Large Audio Language Models excel in general audio understanding, they are limited in temporal reasoning which may hinder their commercial applications and on device deployment. This paper addresses these challenges and limitations in audio temporal reasoning. First, we introduce a data augmentation technique for generating reliable audio temporal questions and answers using an LLM. Second, we propose a continued finetuning curriculum learning strategy to specialize in temporal reasoning without compromising performance on finetuned tasks. Finally, we develop a reliable and transparent automated metric, assisted by an LLM, to measure the correlation between Large Audio Language Model responses and ground truth data intelligently. We demonstrate the effectiveness of our proposed techniques using SOTA LALMs on public audio benchmark datasets.
Title:
Keyword-Aware ASR Error Augmentation for Robust Dialogue State Tracking
Authors: Jihyun Lee, Solee Im, Wonjun Lee, Gary Geunbae Lee
Subjects: Subjects:
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Abstract
Dialogue State Tracking (DST) is a key part of task-oriented dialogue systems, identifying important information in conversations. However, its accuracy drops significantly in spoken dialogue environments due to named entity errors from Automatic Speech Recognition (ASR) systems. We introduce a simple yet effective data augmentation method that targets those entities to improve the robustness of DST model. Our novel method can control the placement of errors using keyword-highlighted prompts while introducing phonetically similar errors. As a result, our method generated sufficient error patterns on keywords, leading to improved accuracy in noised and low-accuracy ASR environments.
Title:
Learning Augmentation Policies from A Model Zoo for Time Series Forecasting
Authors: Haochen Yuan, Xuelin Li, Yunbo Wang, Xiaokang Yang
Abstract
Time series forecasting models typically rely on a fixed-size training set and treat all data uniformly, which may not effectively capture the specific patterns present in more challenging training samples. To address this issue, we introduce AutoTSAug, a learnable data augmentation method based on reinforcement learning. Our approach begins with an empirical analysis to determine which parts of the training data should be augmented. Specifically, we identify the so-called marginal samples by considering the prediction diversity across a set of pretrained forecasting models. Next, we propose using variational masked autoencoders as the augmentation model and applying the REINFORCE algorithm to transform the marginal samples into new data. The goal of this generative model is not only to mimic the distribution of real data but also to reduce the variance of prediction errors across the model zoo. By augmenting the marginal samples with a learnable policy, AutoTSAug substantially improves forecasting performance, advancing the prior art in this field with minimal additional computational cost.
Title:
EntAugment: Entropy-Driven Adaptive Data Augmentation Framework for Image Classification
Authors: Suorong Yang, Furao Shen, Jian Zhao
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Data augmentation (DA) has been widely used to improve the generalization of deep neural networks. While existing DA methods have proven effective, they often rely on augmentation operations with random magnitudes to each sample. However, this approach can inadvertently introduce noise, induce distribution shifts, and increase the risk of overfitting. In this paper, we propose EntAugment, a tuning-free and adaptive DA framework. Unlike previous work, EntAugment dynamically assesses and adjusts the augmentation magnitudes for each sample during training, leveraging insights into both the inherent complexities of training samples and the evolving status of deep models. Specifically, in EntAugment, the magnitudes are determined by the information entropy derived from the probability distribution obtained by applying the softmax function to the model's output. In addition, to further enhance the efficacy of EntAugment, we introduce a novel entropy regularization term, EntLoss, which complements the EntAugment approach. Theoretical analysis further demonstrates that EntLoss, compared to traditional cross-entropy loss, achieves closer alignment between the model distributions and underlying dataset distributions. Moreover, EntAugment and EntLoss can be utilized separately or jointly. We conduct extensive experiments across multiple image classification tasks and network architectures with thorough comparisons of existing DA methods. Importantly, the proposed methods outperform others without introducing any auxiliary models or noticeable extra computational costs, highlighting both effectiveness and efficiency. Code is available at this https URL.
Title:
AMNS: Attention-Weighted Selective Mask and Noise Label Suppression for Text-to-Image Person Retrieval
Authors: Runqing Zhang, Xue Zhou
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Text-to-image person retrieval aims to retrieve images of person given textual descriptions, and most methods implicitly assume that the training image-text pairs are correctly aligned, but in practice, under-correlated and false-correlated problems arise for image-text pairs due to poor image quality and mislabeling. Meanwhile, the random masking augmentation strategy may incorrectly discard semantic content resulting in the problem of generating noisy pairings between image lexical elements and text descriptions. To solve these two problems, we propose a new noise label suppression method and alleviate the problem generated by random mask through an attention-weighted selective mask strategy. In the proposed noise label suppression method, the effect of noise labels is suppressed by preventing the model from being overconfident by considering the inverse KL scatter loss, which is combined with the weight adjustment focus loss to further improve the model's recognition ability on difficult samples. On the other hand, Attention-Weighted Selective Mask processes the raw image through the EMA version of the image encoder, retaining some of the tokens with strong semantic associations with the corresponding text descriptions in order to extract better features. Numerous experiments validate the effectiveness of our approach in terms of dealing with noisy problems. The code will be available soon at this https URL.
Title:
Semi-Supervised 3D Object Detection with Chanel Augmentation using Transformation Equivariance
Authors: Minju Kang, Taehun Kong, Tae-Kyun Kim
Subjects: Subjects:
Computer Vision and Pattern Recognition (cs.CV)
Abstract
Accurate 3D object detection is crucial for autonomous vehicles and robots to navigate and interact with the environment safely and effectively. Meanwhile, the performance of 3D detector relies on the data size and annotation which is expensive. Consequently, the demand of training with limited labeled data is growing. We explore a novel teacher-student framework employing channel augmentation for 3D semi-supervised object detection. The teacher-student SSL typically adopts a weak augmentation and strong augmentation to teacher and student, respectively. In this work, we apply multiple channel augmentations to both networks using the transformation equivariance detector (TED). The TED allows us to explore different combinations of augmentation on point clouds and efficiently aggregates multi-channel transformation equivariance features. In principle, by adopting fixed channel augmentations for the teacher network, the student can train stably on reliable pseudo-labels. Adopting strong channel augmentations can enrich the diversity of data, fostering robustness to transformations and enhancing generalization performance of the student network. We use SOTA hierarchical supervision as a baseline and adapt its dual-threshold to TED, which is called channel IoU consistency. We evaluate our method with KITTI dataset, and achieved a significant performance leap, surpassing SOTA 3D semi-supervised object detection models.
Keyword: detection
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Keyword: face recognition
Title:
Keyword: augmentation
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Title: