Cost-effective and scalable video analytics are essential for precision livestock monitoring, where high-resolution footage and near-real-time monitoring needs from commercial farms generates substantial computational workloads. This paper presents a comprehensive case study on optimizing a poultry welfare monitoring system through system-level improvements across detection, tracking, clustering, and behavioral analysis modules. We introduce a set of optimizations, including multi-level parallelization, Optimizing code with substituting CPU code with GPU-accelerated code, vectorized clustering, and memory-efficient post-processing. Evaluated on real-world farm video footage, these changes deliver up to a 2 × speedup across pipelines without compromising model accuracy. Our findings highlight practical strategies for building high-throughput, low-latency video inference systems that reduce infrastructure demands in agricultural and smart sensing deployments as well as other large-scale video analytics applications.

The integration of Artificial Intelligence (AI) across various fields has transformed the educational landscape and demands a targeted approach to teaching AI in an academic setting. As lecturers aim to prepare students for an AI-driven future, they face various challenges arising from the complex and mathematical nature of AI. This paper explores the challenges of teaching and assessing AI modules in large classrooms by implementing a student-centred approach alongside formative assessment and feedback. It also examines issues related to the diversity of students' skill sets and learning style. This study was conducted on two different cohorts of the same module, Video Analytics and Machine Learning during 2022-2024. Two distinct cohorts were chosen to ensure unbiased conclusions in our study. By recognising and actively addressing these challenges, lecturers can more effectively equip students with the skills needed to navigate this rapidly evolving field. In conclusion, this study shows that implementing formative assessments like quizzes and student-centred approaches are highly beneficial in large classrooms and lead to a significant improvement in student performance and learning outcomes. In addition, the analysis shows that students who are more actively engaged with quizzes tend to score higher on the module. While the overall student feedback has been positive and there has been noticeable improvement in performance, it is important to recognise that there have been instances of unsatisfactory student outcomes as well.

As malware variants continue to increase, the risk of evading detection also grows. While fuzzy hashing has traditionally been successful at clustering malware samples based on their structural similarities, its potential as an active defense tool remains largely unexplored. This study investigates the application of fuzzy hashing for the static analysis of malware binaries to identify common vulnerabilities prevalent in malware, serving as a proactive security measure. We utilized a labeled dataset comprising real-world and synthetic Windows binaries to evaluate six fuzzy hashing algorithms for full binary classification and two for function-level matching. Our results indicate that, while fuzzy hashing is effective in simpler tasks such as malware classification and binary-level vulnerability detection, its performance decreases in more complex scenarios, including multi-class vulnerability identification and matching functions from real-world malware. Moreover, we observed a significant drop in performance, by up to 50%, when transitioning from synthetic to actual malware functions. These findings highlight both the potential and limitations of fuzzy hashing in vulnerability analysis, emphasizing the necessity for more robust techniques to detect vulnerable patterns in real-world malware.

Abstract Many image datasets contain Spurious Correlations (SC), which are coincidental correlations between non-predictive features of the training images and the target label. A classifier trained on such a dataset will appear to perform well when evaluated on the training dataset, but will perform poorly in real-world testing when the spurious correlation is no longer present. This paper investigates the research question of how image classification models can be made robust to the presence of spurious correlations in their training data. To address this challenge, we propose UnLearning from Experience (ULE), a novel student-teacher framework that mitigates SC without requiring group labels. Our method is based on using two classification models trained in parallel: student and teacher models. Both models receive the same batches of training data. The student model is trained with no constraints and pursues the spurious correlations in the data. The teacher model is trained to solve the same classification problem while avoiding the mistakes of the student model. As training is done in parallel, the better the student model learns the spurious correlations, the more robust the teacher model becomes. The teacher model uses the gradient of the student’s output with respect to its input to unlearn mistakes made by the student. Empirically, ULE improves worst-group accuracy by up to 29.0% on Waterbirds, 44.2% on CelebA, 29.4% on Spawrious, and 43.2% on UrbanCars compared to the baseline method.

Abstract Despite the success of Transformer-based language models in a wide variety of natural language processing tasks, our understanding of how these models process a given input in order to represent task-relevant information remains incomplete. In this work, we focus on semantic composition and examine how Transformer-based language models represent semantic information related to the meaning of English noun-noun compounds. We probe Transformer-based language models for their knowledge of the thematic relations that link the head nouns and modifier words of compounds (e.g., KITCHEN CHAIR: a chair located in a kitchen). Firstly, using a dataset featuring groups of compounds with shared lexical or semantic features, we find that token representations of six Transformer-based language models distinguish between pairs of compounds based on whether they use the same thematic relation. Secondly, we utilize fine-grained vector representations of compound semantics derived from human annotations, and find that token vectors from several models elicit a strong signal of the semantic relations used in the compounds. In a novel “compositional probe” setting, where we compare the semantic relation signal in mean-pooled token vectors of compounds to mean-pooled token vectors when the two constituent words appear in separate sentences, we find that the Transformer-based language models that best represent the semantics of noun-noun compounds also do so substantially better than in the control condition where the two constituent works are processed separately. Overall, our results shed light on the ability of Transformer-based language models to support compositional semantic processes in representing the meaning of noun-noun compounds.

In malware analysis, understanding the reasons behind a decision is important for building trust on the system. In the case of opcode-sequence-based classifiers, when standard explanation methods, such as LIME, are applied, the resulting explanation may not provide much insight into the salient parts of the input sequence. This is because LIME treats each opcode as an independent feature, and perturbing this feature will not cause a significant change in the output, meaning the resulting explanation tends to look like random noise. In this paper, we introduce a novel method Hierarchical-LIME (H-LIME) to address this issue. We take into consideration the hierarchical structure of the program, composed of classes and methods. We show that when H-LIME is applied at the level of classes and methods the resulting explanation is sparser, vastly helping improve its interpretability. We conduct extensive experiments by evaluating our proposed method against criteria for accuracy, completeness, sparsity, stability and efficiency. We show that our method significantly improves on all the evaluation criteria compared to other explainability methods.

The global trend of increasing life expectancy introduces new challenges with far-reaching implications. Among these, the risk of falls among older adults is particularly significant, affecting individual health and the quality of life, and placing an additional burden on healthcare systems. Existing fall detection systems often have limitations, including delays due to continuous server communication, high false-positive rates, low adoption rates due to wearability and comfort issues, and high costs. In response to these challenges, this work presents a reliable, wearable, and cost-effective fall detection system. The proposed system consists of a fit-for-purpose device, with an embedded algorithm and an Inertial Measurement Unit (IMU), enabling real-time fall detection. The algorithm combines a Threshold-Based Algorithm (TBA) and a neural network with low number of parameters based on a Transformer architecture. This system demonstrates notable performance with 95.29% accuracy, 93.68% specificity, and 96.66% sensitivity, while only using a 0.38% of the trainable parameters used by the other approach.

Abstract In recent years, the Internet of Things (IoT) has grown rapidly, as has the number of attacks against it. Certain limitations of the paradigm, such as reduced processing capacity and limited main and secondary memory, make it necessary to develop new methods for detecting attacks in real time as it is difficulty to adapt as has the techniques used in other paradigms. In this paper, we propose an architecture capable of generating complex event processing (CEP) rules for real-time attack detection in an automatic and completely unsupervised manner. To this end, CEP technology, which makes it possible to analyze and correlate a large amount of data in real time and can be deployed in IoT environments, is integrated with principal component analysis (PCA), Gaussian mixture models (GMM) and the Mahalanobis distance. This architecture has been tested in two different experiments that simulate real attack scenarios in an IoT network. The results show that the rules generated achieved an F1 score of .9890 in detecting six different IoT attacks in real time.

The Mapper algorithm is a technique from TDA used to create low-dimensional graph-based representations of high-dimensional data, proven effective in numerous exploratory data analysis tasks. The Mapper algorithm’s output depends on several user-chosen parameters, and selecting their values is a non-trivial choice, significantly narrowing its potential application in real-world scenarios. Research attempting to assist in selection of the parameters has been very limited to date. This paper is the first one to address the selection of Mapper’s three parameters simultaneously. The proposed idea incorporates the concept of Ensemble Learning into the Mapper algorithm. Using several datasets with known labels, we show that our method outperforms two baselines in recovering the dataset structure.

Hyper-spectral imaging captures spatial and spectral information of a subject. This is used for the identification of substances within a scene, and food analysis. Presented is an investigation into the capabilities of encoder/decoder deep learning architectures for hyper-spectral image reconstruction from RGB images. For this analysis state-of-the-art (SOTA) techniques for hyper-spectral image reconstruction and other architectures from different fields have been used. Our approach examines a food science case study, using a CPU-based server and different accelerators. An in-house multi-sensor setup was used to capture the dataset which contains hyper-spectral images of twenty slices of different Spanish Ham in the range of 400-100∼nm and their analogous RGB images. The results show no degradation in the output when moving outside of the visual range. This study shows that the SOTA methods for reconstructing from RGB do not produce the most accurate reconstruction of the spectral domain within the range of 400-1000∼nm.

Deep metric learning aims to create a feature space where the projected samples have maximized inter-class and minimized intra-class distance. Most approaches employ only distance-based metrics to achieve this objective, but neglect other properties of the projections in the embedding space, such as their density, sparsity and presence of outliers. In this paper, we propose a novel density-based regulariser, LOFReg, designed to be used as a complement to previously proposed distance-based metric learning loss functions for the re-identification (ReID) and few-shot classification (FSC) tasks. Our method is based on the well-known, in anomaly detection literature, local outlier factor (LOF) algorithm, which estimates the local density deviation of a data point with respect to its neighbours. These measurements are used in our regularisation methodology to achieve an embedding space of evenly distributed samples and to increase the generalisation ability of the model compared to solely distance-based learning. Comprehensive experiments on four publicly available datasets for ReID and FSC, demonstrate consistent improvement against previously proposed metric learning loss functions. Particularly, our experiments show up to 5% improvement in ReID settings, up to 13% in FSC settings and up to 6% improvement against other state-of-art regularisers.

The Internet of Things (IoT) has grown rapidly to become the core of many areas of application, leading to the integration of sensors, with IoT devices. However, the number of attacks against these types of devices has grown as fast as the paradigm itself. Certain inherent characteristics of the paradigm, as well as the limited computational capabilities of the devices involved, make it difficult to deploy security measures. This is why it is necessary to design, implement and study new solutions in the field of cybersecurity. In this paper, we propose an architecture that is capable of generating Complex Event Processing (CEP) rules automatically by integrating them with machine learning technologies. While the former is used to automatically detect attack patterns in real time, the latter, through the use of the Principal Component Analysis (PCA) algorithm, allows the characterization of events and the recognition of anomalies. This combination makes it possible to achieve efficient CEP rules at the computational level, with the results showing that the CEP rules obtained using our approach substantially improve upon the performance of the standard CEP rules, which are rules that are not generated by our proposal but can be defined independently by an expert in the field. Our proposal has achieved an F1-score of 0.98 on average, a 76 percent improvement in throughput over standard CEP rules, and a reduction in the network overhead of 86 percent over standard simple events, which are the simple events that are generated when our proposal is not used.

The adoption of aggressively down-scaled voltages along with worsening process variations, render nanometer devices prone to timing errors that threaten system functionality. The increased vulnerability of nanometer circuits to these errors attracted recent efforts in the development of timing error prediction models using machine learning (ML) methods. However, the majority of such models may be inaccurate, since they either neglect important microarchitecture properties and workload-dependent parameters, affecting timing error manifestation, or are constrained to limited operating areas. In this paper, we propose microarchitecture- and workload-aware ML models for timing error prediction that jointly consider various instruction types as well as all in-flight instructions in a pipeline. Our proposed models are able to predict the exact time and location (i.e., cycle, instruction and bit position) of timing errors with over 98% accuracy across multiple, critical operating regions. To circumvent the increased model complexity due to the considered features, we apply for the first time significance-driven approximations. Evaluation results for various workloads and voltage reduction levels show that our significance-driven precision scaling improves the models’ inference time up to 4.66×, with less than 4% accuracy loss. Finally, we use the proposed model to accurately and realistically inject timing errors during the evaluation of application resiliency. When compared to prior timing error evaluation frameworks that rely on workload-agnostic models, our framework improves the output quality estimation up to 82.6%.

Traditional vehicle re-identification (ReID) approaches, based on visible spectrum data achieve high performance, but have limited capability of real-life applications, as they perform poorly under occluded visibility conditions, such as night-time and bad weather. In such cases, the use of infrared spectrum thermal imagery offers complementary and persistent information when the visual data contribution is inadequate. It is therefore highly beneficial to create a vehicle ReID framework that can exploit both modalities, if available, and is able to apply cross-modality matching, when ReID is required across single modal sensors. This is an extremely challenging task because the nature of the two data modalities induces high discrepancy between the two domains. In this paper we propose a robust end-to-end 2-stream vehicle ReID system that aims to solve the multi-modal and cross-modal ReID problem together by minimising the domain shift between infrared and visible distributions. Our framework consists of a shared network part, following the 2 independent streams, to extract shareable features, along with a domain alignment technique to narrow the gap between the two domains and inter-modality learning to address the cross-domain matching problem. The proposed system achieves state-of-the-art results on RGBN300 dataset, when both modalities are available at inference time. Moreover, our work is the first to explore the cross-modal settings for vehicle ReID and attempts to reduce the performance drop of the cross-modal scenario, when the query and the gallery images come from different modalities. We first measure the baseline cross-modal performance, and then prove that the proposed method improves up to 11% in mAP and 16% in Rank-1 score against the baseline.

In this paper we propose a deep learning based approach for image retrieval using EEG. Our approach makes use of a multi-modal deep neural network based on metric learning, where the EEG signal from a user observing an image is mapped together with visual information extracted from the image. The inspiration behind this work is the vision of a system which allows the user to navigate their image catalogue just by thinking about the image they want to see. Thanks to our metric learning approach, the system is scalable in that it can operate with new images that have never been used in training, resulting in a zero-shot image retrieval system. This framework is tested in two different standard EEG image-viewing datasets, where we demonstrate state-of-the-art results in this complex scenario.

We propose a new method for single-camera real-world 3-D human pose estimation. Our method uses multitask training together with iterative pose refinement using a novel conditional attention mechanism. For iterative pose refinement, the output of each convolutional layer is conditioned on the latest pose estimate, using a conditioned squeeze-and-excitation network architecture that incorporates novel feedback connections. Multitask training on both an in-the-wild 2-D pose dataset and a controlled 3-D pose dataset allows for real-world 3-D pose estimation without the need for a large-scale in-the-wild 3-D pose dataset, which is unavailable. Experiments are performed on several real-world datasets, as well as the Human 3.6 Million and HumanEva-I datasets, to show that the combined attention mechanism, iterative refinement scheme, and multitask training allow us to achieve robust and competitive performance with only a simple network architecture. In addition, we show that our method is efficient enough to run on commodity hardware, producing pose estimates in real time.

We present DANdroid, a novel Android malware detection model using a deep learning Discriminative Adversarial Network (DAN) that classifies both obfuscated and unobfuscated apps as either malicious or benign. Our method, which we empirically demonstrate is robust against a selection of four prevalent and real-world obfuscation techniques, makes three contributions. Firstly, an innovative application of discriminative adversarial learning results in malware feature representations with a strong degree of resilience to the four obfuscation techniques. Secondly, the use of three feature sets; raw opcodes, permissions and API calls, that are combined in a multi-view deep learning architecture to increase this obfuscation resilience. Thirdly, we demonstrate the potential of our model to generalize over rare and future obfuscation methods not seen in training. With an overall dataset of 68,880 obfuscated and unobfuscated malicious and benign samples, our multi-view DAN model achieves an average F-score of 0.973 that compares favourably with the state-of-the-art, despite being exposed to the selected obfuscation methods applied both individually and in combination.

Identifying vehicles across a network of cameras with non-overlapping fields of view remains a challenging research problem due to scene occlusions, significant inter-class similarity and intra-class variability. In this paper, we propose an end-to-end multi-level re-identification network that is capable of successfully projecting same identity vehicles closer to one another in the embedding space, compared to vehicles of different identities. Robust feature representations are obtained by combining features at multiple levels of the network. As for the learning process, we employ a recent state-of-the-art structured metric learning loss function previously applied to other retrieval problems and adjust it to the vehicle re-identification task. Furthermore, we explore the cases of image-to-image, image-to-video and video-to-video similarity metric. Finally, we evaluate our system and achieve great performance on two large-scale publicly available datasets, CityFlow-ReID and VeRi-776. Compared to most existing state-of-art approaches, our approach is simpler and more straightforward, utilizing only identity-level annotations, while avoiding post-processing the ranking results (re-ranking) at the testing phase.

Distributed Denial of Service (DDoS) attacks are one of the most harmful threats in today's Internet, disrupting the availability of essential services. The challenge of DDoS detection is the combination of attack approaches coupled with the volume of live traffic to be analysed. In this paper, we present a practical, lightweight deep learning DDoS detection system called Lucid, which exploits the properties of Convolutional Neural Networks (CNNs) to classify traffic flows as either malicious or benign. We make four main contributions; (1) an innovative application of a CNN to detect DDoS traffic with low processing overhead, (2) a dataset-agnostic preprocessing mechanism to produce traffic observations for online attack detection, (3) an activation analysis to explain Lucid's DDoS classification, and (4) an empirical validation of the solution on a resource-constrained hardware platform. Using the latest datasets, Lucid matches existing state-of-the-art detection accuracy whilst presenting a 40x reduction in processing time, as compared to the state-of-the-art. With our evaluation results, we prove that the proposed approach is suitable for effective DDoS detection in resource-constrained operational environments.

Electromyography (EMG) signal analysis is a popular method for controlling prosthetic and gesture control equipment. For portable systems, such as prosthetic limbs, real-time low-power operation on embedded processors is critical, but to date there has been no record of how existing EMG analysis approaches support such deployments. This paper presents a novel approach to time-domain classification of multichannel EMG signals harnessed from randomly-placed sensors according to the wrist-hand movements which caused their occurrence. It shows how, by employing a very small set of time-domain features, Kernel Fisher discriminant feature projection and Radial Bias Function neural network classifiers, nine wrist-hand movements can be detected with accuracy exceeding 99% - surpassing the state-of-the-art on record. It also shows how, when deployed on ARM Cortex-A53, the processing time is not only sufficient to enable real-time processing but is also a factor 50 shorter than the leading time-frequency techniques on record.

In this paper, we propose a video-based person re-identification system for wide area tracking based on a recurrent neural network architecture. Given short video sequences of a person, generated by a tracking algorithm, our video re-identification algorithm links these tracklets in full trajectories across a network of non-overlapping cameras in an open-world scenario. In our system, features are first extracted from each frame using a convolutional neural network. Then, a recurrent layer combines information across time-steps. The features from all time-steps are finally combined using temporal pooling to give an overall appearance feature for the complete sequence. Our system is trained to perform re-identification using a Siamese network architecture. Experiments are conducted on the iLIDS-VID and PRID-2011 video re-identification data sets as well as in the DukeMTMC multi-camera tracking data set.

This paper presents a unique real-time motion recognition system for Electromyographic (EMG) signal acquisition and classification. It is the first approach which can classify hand poses from multi-channel EMG signals gathered from randomly placed arm sensors as accurately as current placed-sensor EMG acquisition approaches. It combines time-domain feature extraction, Linear Discriminant Analysis (LDA) feature projection and Multilayer Perceptron (MLP) classification to allow nine distinct poses to be correctly identified more than 95% of the time. This is comparable to state-of-the-art placed-sensor EMG acquisition systems. Processing times of 11.70 ms also make this a viable candidate approach for real-time EMG acquisition and processing in practical prosthesis applications.

In this paper, we propose a novel android malware detection system that uses a deep convolutional neural network (CNN). Malware classification is performed based on static analysis of the raw opcode sequence from a disassembled program. Features indicative of malware are automatically learned by the network from the raw opcode sequence thus removing the need for hand-engineered malware features. The training pipeline of our proposed system is much simpler than existing n-gram based malware detection methods, as the network is trained end-to-end to jointly learn appropriate features and to perform classification, thus removing the need to explicitly enumerate millions of n-grams during training. The network design also allows the use of long n-gram like features, not computationally feasible with existing methods. Once trained, the network can be efficiently executed on a GPU, allowing a very large number of files to be scanned quickly.

Person reidentification involves recognizing a person across nonoverlapping camera views, with different pose, illumination, and camera characteristics. We propose to tackle this problem by training a deep convolutional network to represent a person's appearance as a low-dimensional feature vector that is invariant to common appearance variations encountered in the reidentification problem. Specifically, a Siamese network architecture is used to train a feature extraction network using pairs of similar and dissimilar images. We show that the use of a novel multitask learning objective is crucial for regularizing the network parameters in order to prevent overfitting due to the small size of the training data set. We complement the verification task, which is at the heart of reidentification, by training the network to jointly perform verification and identification and to recognize attributes related to the clothing and pose of the person in each image. In addition, we show that our proposed approach performs well even in the challenging cross-data set scenario, which may better reflect real-world expected performance.

In this paper we propose a novel recurrent neural network architecture for video-based person re-identification. Given the video sequence of a person, features are extracted from each frame using a convolutional neural network that incorporates a recurrent final layer, which allows information to flow between time-steps. The features from all timesteps are then combined using temporal pooling to give an overall appearance feature for the complete sequence. The convolutional network, recurrent layer, and temporal pooling layer, are jointly trained to act as a feature extractor for video-based re-identification using a Siamese network architecture. Our approach makes use of colour and optical flow information in order to capture appearance and motion information which is useful for video re-identification. Experiments are conduced on the iLIDS-VID and PRID-2011 datasets to show that this approach outperforms existing methods of video-based re-identification.

In this paper we explore ways to address the issue of dataset bias in person re-identification by using data augmentation to increase the variability of the available datasets, and we introduce a novel data augmentation method for re-identification based on changing the image background. We show that use of data augmentation can improve the cross-dataset generalisation of convolutional network based re-identification systems, and that changing the image background yields further improvements.

In this paper we extend the minimum-cost network flow approach to multi-target tracking, by incorporating a motion model, allowing the tracker to better cope with long term occlusions and missed detections. In our new method, the tracking problem is solved iteratively: Firstly, an initial tracking solution is found without the help of motion information. Given this initial set of track lets, the motion at each detection is estimated, and used to refine the tracking solution. Finally, special edges are added to the tracking graph, allowing a further revised tracking solution to be found, where distant track lets may be linked based on motion similarity. Our system has been tested on the PETS S2.L1 and Oxford town-center sequences, outperforming the baseline system, and achieving results comparable with the current state of the art.

We introduce a novel dual-stage algorithm for online multitarget tracking in realistic conditions. In the first stage, the problem of data association between tracklets and detections, given partial occlusion, is addressed using a novel occlusion robust appearance similarity method. This is used to robustly link tracklets with detections without requiring explicit knowledge of the occluded regions. In the second stage, tracklets are linked using a novel method of constraining the linking process that removes the need for ad-hoc tracklet linking rules. In this method, links between tracklets are permitted based on their agreement with optical flow evidence. Tests of this new tracking system have been performed using several public datasets.

This paper presents generalized Laplacian eigenmaps, a novel dimensionality reduction approach designed to address stylistic variations in time series. It generates compact and coherent continuous spaces whose geometry is data-driven. This paper also introduces graph-based particle filter, a novel methodology conceived for efficient tracking in low dimensional space derived from a spectral dimensionality reduction method. Its strengths are a propagation scheme, which facilitates the prediction in time and style, and a noise model coherent with the manifold, which prevents divergence, and increases robustness. Experiments show that a combination of both techniques achieves state-of-the-art performance for human pose tracking in underconstrained scenarios.

We address the problem of multi-target tracking in realistic crowded conditions by introducing a novel dual-stage online tracking algorithm. The problem of data-association between tracks and detections, based on appearance, is often complicated by partial occlusion. In the first stage, we address the issue of occlusion with a novel method of robust data-association, that can be used to compute the appearance similarity between tracks and detections without the need for explicit knowledge of the occluded regions. In the second stage, broken tracks are linked based on motion and appearance, using an online-learned linking model. The online-learned motion-model for track linking uses the confident tracks from the first stage tracker as training examples. The new approach has been tested on the town centre dataset and has performance comparable with the present state-of-the-art.

A bstract In this paper, we propose an integrated particle filter-based pose tracking framework which combines priors able to model human motions keeping stylistic variations, reducing the probability of divergence and facilitating the recovering after failure. A novel unsupervised dimensionality reduction technique, Generalised Laplacian Eigenmaps (GLE), generates compact and coherent continuous spaces which explicitly express style. The proposed particle filter embeds the GLE manifold to take advantage of its geometry into the propagation and hypothesis generation stage. The method is validated using standard HumanEva 2 dataset.

This paper proposes a two-level 3D human pose tracking method for a specific action captured by several cameras. The generation of pose estimates relies on fitting a 3D articulated model on a Visual Hull generated from the input images. First, an initial pose estimate is constrained by a low dimensional manifold learnt by Temporal Laplacian Eigenmaps. Then, an improved global pose is calculated by refining individual limb poses. The validation of our method uses a public standard dataset and demonstrates its accurate and computational efficiency.

In this paper, a novel framework for visual tracking of human body parts is introduced. The approach presented demonstrates the feasibility of recovering human poses with data from a single uncalibrated camera by using a limb-tracking system based on a 2-D articulated model and a double-tracking strategy. Its key contribution is that the 2-D model is only constrained by biomechanical knowledge about human bipedal motion, instead of relying on constraints that are linked to a specific activity or camera view. These characteristics make our approach suitable for real visual surveillance applications. Experiments on a set of indoor and outdoor sequences demonstrate the effectiveness of our method on tracking human lower body parts. Moreover, a detail comparison with current tracking methods is presented.

A novel non-linear dimensionality reduction method, called Temporal Laplacian Eigenmaps, is introduced to process efficiently time series data. In this embedded-based approach, temporal information is intrinsic to the objective function, which produces description of low dimensional spaces with time coherence between data points. Since the proposed scheme also includes bidirectional mapping between data and embedded spaces and automatic tuning of key parameters, it offers the same benefits as mapping-based approaches. Experiments on a couple of computer vision applications demonstrate the superiority of the new approach to other dimensionality reduction method in term of accuracy. Moreover, its lower computational cost and generalisation abilities suggest it is scalable to larger datasets.

In this paper, a novel framework for visual tracking of human body parts is introduced. The presented approach demonstrates the feasibility of recovering human poses with data from a single uncalibrated camera using a limb tracking system based on a 2D articulated model. It is constrained only by biomechanical knowledge about human bipedal motion, instead on relying on constraints linked to a specific activity or camera view. These characteristics make our approach suitable for real visual surveillance applications. Experiments on HumanEva I & II datasets demonstrate the effectiveness of our method on tracking human lower body parts. Moreover, a detail comparison with current tracking methods is presented. 1

In this paper, we introduce an efficient method for particle selection in tracking objects in complex scenes. First, we improve the proposal distribution function of the tracking algorithm, including current observation, reducing the cost of evaluating particles with a very low likelihood. In addition, we use a partitioned sampling approach to decompose the dynamic state in several stages. It enables to deal with high-dimensional states without an excessive computational cost. To represent the color distribution, the appearance of the tracked object is modelled by sampled pixels. Based on this representation, the probability of any observation is estimated using non-parametric techniques in color space. As a result, we obtain a probability color density image (PDI) where each pixel points its membership to the target color model. In this way, the evaluation of all particles is accelerated by computing the likelihood p(z\x) using the integral image of the PDI.

This work addresses the problem of human motion analysis in video sequences of a scene observed by a single fixed camera with high perspective effect. The goal of this work is to make a 2D-Model (made of Shape and Stick figure) viewpoint-insensitive and preprocess the input image for removing the perspective effect. We focus our methodology on using the 3D principal directions of man-made environments and also the direction of motion to transform both 2D-Model and input images to a common frontal view (parallel or orthogonal to the direction of motion) before the fitting process. The inverse transformation is then performed on the resulting human features obtaining a segmented silhouette and a pose estimation in the original input image. Preliminary results are very promising since the proposed algorithm is able to locate head and feet with a better precision than previous one. 1