Chen Feng

Technological advances have made VR interactions more natural, yet fatigue and social awkwardness persist. Microgestures offer a promising solution. However, prior research has focused mainly on the index finger, leaving the thumb’s potential as a stable, low-fatigue ray-casting modality underexplored. This study conducted two experiments to examine whether the thumb can function as an effective technique for 3D pointing in VR. Experiment 1 compared four input methods: VR controller, hand tracking, index-finger pointing, and thumb pointing. The controller was most efficient, while hand tracking provided greater stability but higher fatigue. Both microgestures outperformed hand tracking in efficiency and fatigue reduction. Experiment 2 examined thumb and index-finger pointing on horizontal and vertical target planes. Vertical layouts facilitated faster and more efficient pointing. The index finger was more efficient overall, but the thumb caused slightly less fatigue during vertical interactions. These findings provide valuable guidance for the design of microgesture-based VR interaction.

Museums increasingly rely on digital content to support visitors’ understanding of artworks, yet little is known about how these formats shape the emotional engagement that underlies meaningful art experiences. This research presents an in-situ EEG study on how digital interpretive content modulate engagement during art viewing. Participants experienced three modalities: direct viewing of a Bruegel painting, a 180° immersive interpretive projection, and a regular, display-based interpretive video. Frontal EEG markers of motivational orientation, internal involvement, perceptual drive, and arousal were extracted using eyes-open baselines and Z-normalized contrasts. Results show modality-specific engagement profiles: display-based interpretive video induced high arousal and fast-band activity, immersive projections promoted calm, presence-oriented absorption, and original artworks reflected internally regulated engagement. These findings, relying on lightweight EEG sensing in an operational cultural environment, suggest that digital interpretive content affects engagement style rather than quantity. This paves the way for new multimodal sensing approaches and enables museums to optimize the modalities and content of their interpretive media.

Objective To investigate the intervention effect of digital music meditation on anxiety and depression of college students. Methods 80 college students with significant anxiety or depression (SAS≥50 or SDS≥53) were randomly assigned to group A and group B. Two sets of digital music meditation audio tools were used for continuous 4 weeks of meditation intervention. After 4 weeks, the two Self-rating scales (SAS and SDS) were used again to evaluate the degree of anxiety and depression. Results After four weeks of intervention, the scoring of anxiety and depression was significantly lower than those before intervention (SAS: <italic>t</italic>=11.160, <italic>P</italic>&lt;.05; SDS: <italic>t</italic>=12.603, <italic>P</italic>&lt;.05), there was a significant difference in the decline rate of anxiety and depression scores (<italic>t</italic>=-3.219, <italic>P</italic>=.002). There was no significant difference in anxiety and depression scores between group A and group B after intervention. Conclusion Two sets of different types of digital music meditation audio tools can significantly improve the anxiety and depression of college students in group A and B during the epidemic, but there is no significant difference between the two groups. The effect of digital music meditation on depression was better than that of anxiety.

OBJECTIVE: Postoperative atrial fibrillation (POAF) is a common complication following off-pump coronary artery bypass grafting (CABG). This study aims to assess whether elevated preoperative N-terminal pro-brain natriuretic peptide (NT-proBNP) levels can effectively stratify patients based on their risk of developing POAF. METHODS: Utilizing a retrospective database of 512 patients who underwent off-pump CABG, we compared preoperative clinical data, including NT-proBNP levels, between patients experiencing POAF lasting longer than 30 s during hospitalization and those who did not. RESULTS: POAF manifested in 23.6% of patients (121 out of 512). After off-pump CABG, 39% of patients (100 out of 256) with NT-proBNP levels greater than the median (388 pg/mL) developed POAF, in contrast to only 8% of patients (21 out of 256) with levels below 388 pg/mL (P < 0.0001). NT-proBNP levels were significantly higher in patients with POAF compared to those without (median, 1149 vs. 278 pg/mL; P < 0.0001). Multivariate logistic regression analysis revealed that in patients undergoing off-pump CABG, a NT-proBNP level of 388 pg/mL or greater (OR, 5.72; 95% CI, 3.24-10.48; P < 0.0001) was the only independent risk factor for POAF. CONCLUSIONS: For patients undergoing off-pump CABG, increased age, preoperative left ventricular ejection fraction (LVEF) and left atrial diameter (LAD), creatinine levels, and a preoperative NT-proBNP level of 388 pg/mL or greater emerged as significant risk factors for POAF. Identifying individuals predisposed to POAF enables the design of trials evaluating preventive strategies to mitigate this complication.

Visual localization, which estimates a camera's pose within a known scene, is a fundamental capability for autonomous systems. While absolute pose regression (APR) methods have shown promise for efficient inference, they often struggle with generalization. Recent approaches attempt to address this through data augmentation with varied viewpoints, yet they overlook a critical factor: appearance diversity. In this work, we identify appearance variation as the key to robust localization. Specifically, we first lift real 2D images into 3D Gaussian Splats with varying appearance and deblurring ability, enabling the synthesis of diverse training data that varies not just in poses but also in environmental conditions such as lighting and weather. To fully unleash the potential of the appearance-diverse data, we build a two-branch joint training pipeline with an adversarial discriminator to bridge the syn-to-real gap. Extensive experiments demonstrate that our approach significantly outperforms state-of-the-art methods, reducing translation and rotation errors by 50% and 33% on indoor datasets, and 38% and 44% on outdoor datasets. Most notably, our method shows remarkable robustness in dynamic driving scenarios under varying weather conditions and in day-to-night scenarios, where previous APR methods fail.

It is well known that high dynamic range (HDR) videos enhance immersive visual experiences compared to conventional standard dynamic range content. However, HDR content is typically more challenging to encode due to the increased detail associated with the wider dynamic range. In this work, we improve HDR compression performance using an Effective Bit Depth Adaptation approach (EBDA), which reduces the effective bit depth of the original video content before encoding and reconstructs the full bit depth using a CNN-based up-sampling method at the decoder. The up-sampling deep network is based on a new version of Multi-frame MFRNet, MF-MFRNet. This approach has been integrated into the EBDA framework with two Versatile Video Coding (VVC) reference models: VTM 16.2 and the Fraunhofer Versatile Video Encoder (VVenC 1.4.0). The proposed approach has been evaluated under the JVET HDR Common Test Conditions using the Random Access configuration. The results show evident coding gains over both the original VTM 16.2 and VVenC 1.4.0 on all JVET HDR tested sequences, with average bitrate savings of 3.1% and 4.8% based on PSNR and 7.8% and 9.6% based on VMAF against VTM and VVenC respectively. The source code of multi-frame MFRNet has been released at https://github.com/fan-aaron-zhang/MF-MFRNet.

Lymph node (LN) metastasis is a prevalent cause of recurrence in oral squamous cell carcinoma (OSCC). However, accurately identifying metastatic LNs (LNs+) remains challenging. This prospective clinical study aims to test the effectiveness of our convolutional neural network (CNN) model for identifying OSCC cervical LN+ in contrast-enhanced computed tomography (CECT) in clinical practice. A CNN model was developed and trained using a dataset of 8,380 CECT images from previous OSCC patients. It was then prospectively validated on 17,777 preoperative CECT images from 354 OSCC patients between October 17, 2023, and August 31, 2024. The model’s predicted LN results were provided to the surgical team without influencing surgical or treatment plans. During surgery, the predicted LN+ were identified and sent for separate pathological examination. The accuracy of the model’s predictions was compared with those of human experts and verified against pathology reports. The capacity of the model to assist radiologists in LN+ diagnosis was also assessed. The CNN model was trained over 40 epochs and successfully validated after each. Compared with human experts (2 radiologists, 2 surgeons, and 2 students), the CNN model achieved higher sensitivity (81.89% vs. 81.48%, 46.91%, 50.62%), specificity (99.31% vs. 99.15%, 98.36%, 96.27%), LN+ accuracy (76.19% vs. 75.43%, P = 0.854; 40.64%, P &lt; 0.001; 37.44%, P &lt; 0.001), and clinical accuracy (86.16% vs. 83%, 61%, 56%). With the model’s assistance, the radiologists surpassed both the previous predictive results without the model’s support and the model’s performance alone. The CNN model demonstrated an accuracy comparable to that of radiologists in identifying, locating, and predicting cervical LN+ in OSCC patients. Furthermore, the model has the potential to assist radiologists in making more accurate diagnoses.

<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Goal:</i> Persons with blindness or low vision (pBLV) face challenges in completing activities of daily living (ADLs/IADLs). Semantic segmentation techniques on smartphones, like DeepLabV3+, can quickly assist in identifying key objects, but their performance across different indoor settings and lighting conditions remains unclear. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Methods:</i> Using the MIT ADE20K SceneParse150 dataset, we trained and evaluated AI models for specific indoor scenes (kitchen, bedroom, bathroom, living room) and compared them with a generic indoor model. Performance was assessed using mean accuracy and intersection-over-union metrics. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Results:</i> Scene-specific models outperformed the generic model, particularly in identifying ADL/IADL objects. Models focusing on rooms with more unique objects showed the greatest improvements (bedroom, bathroom). Scene-specific models were also more resilient to low-light conditions. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Conclusions:</i> These findings highlight how using scene-specific models can boost key performance indicators for assisting pBLV across different functional environments. We suggest that a dynamic selection of the best-performing models on mobile technologies may better facilitate ADLs/IADLs for pBLV.

Objective: To establish and validate a model based on hyperdense middle cerebral artery sign (HMCAS) radiomics features for predicting hemorrhagic transformation (HT) in patients with acute ischemic stroke (AIS) after endovascular treatment (EVT). Methods: Patients with AIS who presented with HMCAS on non-contrast computed tomography (NCCT) at admission and underwent EVT at three comprehensive hospitals between June 2020 and January 2024 were recruited for this retrospective study. A radiomics model was constructed using the HMCAS radiomics features most strongly associated with HT. In addition, clinical and radiological independent factors associated with HT were identified. Subsequently, a combined model incorporating radiomics features and independent risk factors was developed via multivariate logistic regression and presented as a nomogram. The models were evaluated via receiver operating characteristic curve, calibration curve, and decision curve analysis. Results: Of the 118 patients, 71 (60.17%) developed HT. The area under the curve (AUC) of the radiomics model was 0.873 (95% CI 0.797-0.935) in the training cohort and 0.851 (95%CI 0.721-0.942) in the test cohort. The Alberta Stroke Program Early CT score (ASPECTS) was the only independent predictor among 24 clinical and 4 radiological variables. The combined model further improved the predictive performance, with an AUC of 0.911 (95%CI 0.850-0.960) in the training cohort and 0.877 (95%CI 0.753-0.960) in the test cohort. Decision curve analysis demonstrated that the combined model had greater clinical utility for predicting HT. Conclusion: HMCAS-based radiomics is expected to be a reliable tool for predicting HT risk stratification in AIS patients after EVT.

Visual place recognition (VPR) using deep networks has achieved state-of-the-art performance. However, most of them require a training set with ground truth sensor poses to obtain positive and negative samples of each observation's spatial neighborhood for supervised learning. When such information is unavailable, temporal neighborhoods from a sequentially collected data stream could be exploited for self-supervised training, although we find its performance suboptimal. Inspired by noisy label learning, we propose a novel self-supervised framework named TF-VPR that uses temporal neighborhoods and learnable feature neighborhoods to discover unknown spatial neighborhoods. Our method follows an iterative training paradigm which alternates between: (1) representation learning with data augmentation, (2) positive set expansion to include the current feature space neighbors, and (3) positive set contraction via geometric verification. We conduct auto-labeling and generalization tests on both simulated and real datasets, with either RGB images or point clouds as inputs. The results show that our method outperforms self-supervised baselines in recall rate, robustness, and heading diversity, a novel metric we propose for VPR.

This paper provides an introduction to the concepts of fitness landscapes and information landscapes(IL), and proposes an improvement to the IL measure, transforming it into an indicator for evaluating the ruggedness of landscapes. A novel extremum-solving algorithm based on the IL measure is introduced for the extremum seeking of high-dimensional functions. The algorithm incorporates Latin hypercube sampling (LHS) from spatial filling designs to sample high-dimensional spaces, thereby obtaining regional sample points in high-dimensional spaces. It then applies the IL measure to calculate intervals, facilitating the extremum solving process of the algorithm. Finally, by comparing this algorithm with other methods, it is demonstrated that the algorithm can find all solutions with fewer initial points, outperforming traditional methods in both speed and solution quality, thus providing a theoretical foundation for subsequent engineering applications.

Visual place recognition (VPR) enables autonomous robots to identify previously visited locations, which contributes to tasks like simultaneous localization and mapping (SLAM). VPR faces challenges such as accurate image neighbor retrieval and appearance change in scenery. Event cameras, also known as dynamic vision sensors, are a new sensor modality for VPR and offer a promising solution to the challenges with their unique attributes: high temporal resolution (1MHz clock), ultra-low latency (in μs), and high dynamic range (&gt;120dB). These attributes make event cameras less susceptible to motion blur and more robust in variable lighting conditions, making them suitable for addressing VPR challenges. However, the scarcity of event-based VPR datasets, partly due to the novelty and cost of event cameras, hampers their adoption. To fill this data gap, our paper introduces the NYC-Event-VPR dataset to the robotics and computer vision communities, featuring the Prophesee IMX636 HD event sensor (1280x720 resolution), combined with RGB camera and GPS module. It encompasses over 13 hours of geotagged event data, spanning 260 kilometers across New York City, covering diverse lighting and weather conditions, day/night scenarios, and multiple visits to various locations. Furthermore, our paper employs three frameworks to conduct generalization performance assessments, promoting innovation in event-based VPR and its integration into robotics applications.

Adopting omnidirectional Field of View (FoV) cameras in aerial robots vastly improves perception ability, significantly advancing aerial robotics’s capabilities in inspection, reconstruction, and rescue tasks. However, such sensors also elevate system complexity, e.g., hardware design, and corresponding algorithm, which limits researchers from utilizing aerial robots with omnidirectional FoV in their research. To bridge this gap, we propose OmniNxt, a fully open-source aerial robotics platform with omnidirectional perception. We design a high-performance flight controller Nxt-FC and a multi-fisheye camera set for OmniNxt. Meanwhile, the compatible software is carefully devised, which empowers OmniNxt to achieve accurate localization and real-time dense mapping with limited computation resource occupancy. We conducted extensive real-world experiments to validate the superior performance of OmniNxt in practical applications. All the hardware and software are open-access at<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup>, and we provide docker images of each crucial module in the proposed system. Project page: https://hkust-aerial-robotics.github.io/OmniNxt.

Unmanned Aerial Vehicles (UAVs) have gained significant popularity in scene reconstruction. This paper presents SOAR, a LiDAR-Visual heterogeneous multi-UAV system specifically designed for fast autonomous reconstruction of complex environments. Our system comprises a LiDAR-equipped explorer with a large field-of-view (FoV), alongside photographers equipped with cameras. To ensure rapid acquisition of the scene’s surface geometry, we employ a surface frontier-based exploration strategy for the explorer. As the surface is progressively explored, we identify the uncovered areas and generate viewpoints incrementally. These viewpoints are then assigned to photographers through solving a Consistent Multiple Depot Multiple Traveling Salesman Problem (Consistent-MDMTSP), which optimizes scanning efficiency while ensuring task consistency. Finally, photographers utilize the assigned viewpoints to determine optimal coverage paths for acquiring images. We present extensive benchmarks in the realistic simulator, which validates the performance of SOAR compared with classical and state-of-the-art methods. For more details, please see our project page at sysu-star.github.io/SOAR.

Understanding passenger travel patterns is helpful for the allocation of passenger resources in urban rail transit. Based on rail transit smart card data, this paper proposes a method to identify travel patterns by modeling individual spatiotemporal sequences. First, all the stations visited by individual passengers are extracted, and the similarity of the stations is calculated by the frequency of inter-station travel, the distance between stations and the activity duration of the stations. The main spatial activity area of the individual is divided using a hierarchical clustering algorithm. Then, the spatiotemporal sequence is inferred based on the individual's travel order. The sequence is a set of discrete values that characterize the spatiotemporal state. PCA-KL and K-Means++ are used to extract the similarity sequence structure to identify the passenger travel pattern. Finally, the rail transit smart card data of Xi'an in a certain month is taken as an example to identify its passenger travel pattern. The results show that complex passenger flow has 5 travel modes, of which 3 typical modes are commuting travel in a macro sense, accounting for 79% of the passenger flow. It can be seen that the pattern recognition based on the similarity of individual spatiotemporal sequences in this paper fully reflects the particularity and versatility of the research method, and is highly operational for different cities.

The industrial world is shifting from mass production to meeting individual needs, requiring guidelines to incorporate emotions and preferences. Traditional questionnaires fall short in capturing detailed real-time emotion changes. Therefore, we propose a novel approach using Electroencephalography(EEG), which reflects central nervous system activity, and heart rate variability(HRV), which reflects autonomic nervous system activity, to estimate emotions in real-time by analyzing arousal and valence. Based on the 2D Arousal-Valence model, we plot EEG and HRV data onto this model to visualize emotion changes caused by external stimuli. Using this proposed method, we conducted joint research with the home and personal care sector to evaluate emotional responses to aroma products, and with the automotive sector to assess drivers' emotional states. These studies demonstrate the value of academic research in guiding the industrial sector, particularly through emotion estimation based on EEG and HRV. Our findings suggest that these technologies offer valuable insights into consumer emotions, which can help improve product design and user experience. Ongoing collaboration between academia and the industrial sector is crucial for overcoming challenges in interpreting physiological signals and effectively integrating these methods into practice.

Action segmentation has made significant progress, but segmenting and recognizing actions from untrimmed long videos remains a challenging problem. Most state-of-the-art methods focus on designing models based on temporal convolution. However, the limitations of modeling long-term temporal dependencies and the inflexibility of temporal convolutions restrict the potential of these models. To address the issue of over-segmentation in existing action segmentation methods, which leads to classification errors and reduced segmentation quality, this paper proposes a global spatial-temporal information encoder-decoder based action segmentation method. The method proposed in this paper uses the global temporal information captured by refinement layer to assist the Encoder-Decoder (ED) structure in judging the action segmentation point more accurately and, at the same time, suppress the excessive segmentation phenomenon caused by the ED structure. The method proposed in this paper achieves 93% frame accuracy on the constructed real Tai Chi action dataset. The experimental results prove that this method can accurately and efficiently complete the long video action segmentation task.

Unmanned Aerial Vehicles (UAVs) have gained significant popularity in scene reconstruction. This paper presents SOAR, a LiDAR-Visual heterogeneous multi-UAV system specifically designed for fast autonomous reconstruction of complex environments. Our system comprises a LiDAR-equipped explorer with a large field-of-view (FoV), alongside photographers equipped with cameras. To ensure rapid acquisition of the scene's surface geometry, we employ a surface frontier-based exploration strategy for the explorer. As the surface is progressively explored, we identify the uncovered areas and generate viewpoints incrementally. These viewpoints are then assigned to photographers through solving a Consistent Multiple Depot Multiple Traveling Salesman Problem (Consistent-MDMTSP), which optimizes scanning efficiency while ensuring task consistency. Finally, photographers utilize the assigned viewpoints to determine optimal coverage paths for acquiring images. We present extensive benchmarks in the realistic simulator, which validates the performance of SOAR compared with classical and state-of-the-art methods. For more details, please see our project page at https://sysu-star.github.io/SOAR}{sysu-star.github.io/SOAR.

UNav is a computer-vision-based localization and navigation aid that provides step-by-step route instructions to reach selected destinations without any infrastructure in both indoor and outdoor environments. Despite the initial literature highlighting UNav's potential, clinical efficacy has not yet been rigorously evaluated. Herein, we assess UNav against standard in-person travel directions (SIPTD) for persons with blindness or low vision (PBLV) in an ecologically valid environment using a non-inferiority design. Twenty BLV subjects (age = 38 ± 8.4; nine females) were recruited and asked to navigate to a variety of destinations, over short-range distances (<200 m), in unfamiliar spaces, using either UNav or SIPTD. Navigation performance was assessed with nine dependent variables to assess travel confidence, as well as spatial and temporal performances, including path efficiency, total time, and wrong turns. The results suggest that UNav is not only non-inferior to the standard-of-care in wayfinding (SIPTD) but also superior on 8 out of 9 metrics, as compared to SIPTD. This study highlights the range of benefits computer vision-based aids provide to PBLV in short-range navigation and provides key insights into how users benefit from this systematic form of computer-aided guidance, demonstrating transformative promise for educational attainment, gainful employment, and recreational participation.

Object detection models available on smartphones such as YOLOv8 can potentially help identify and locate objects of interest to people who are blind or low vision (pBLV). However, current models may miss crucial objects for pBLV. Here, we compared 5 transfer learning methods for adding new classes of interest to pBLV navigation that are absent from the Common Objects in Context (COCO) training dataset. Using a rebalanced COCO dataset with these new classes, we revised public YOLOv8s models via the following methods: revising all pretrained weights; freezing 22, 21 or 15 layers; and few-shot learning. These approaches achieved overall mean average precision (mAP-50) from 0.342 (20 min training time; few-shot learning) to 0.420 (9.2 hrs; revising all pretrained weights). Among the frozen layer models, the 15 frozen layer model had the best mAP-50 performance of 0.419 (7.7 hrs); hyperparameter tuning on this model increased mAP-50 to 0.423. When applied to a larger YOLOv8xl model, mAP-50 reached 0.511 after 50 epochs. Our results highlight how object detection models can be adapted for the benefit of pBLV users even when developers have limited training data or computational resources.

3D coverage path planning for UAVs is a crucial problem in diverse practical applications. However, existing methods have shown unsatisfactory system simplicity, computation efficiency, and path quality in large and complex scenes. To address these challenges, we propose FC-Planner, a skeleton-guided planning framework that can achieve fast aerial coverage of complex 3D scenes without pre-processing. We decompose the scene into several simple subspaces by a skeleton-based space decomposition (SSD). Additionally, the skeleton guides us to effortlessly determine free space. We utilize the skeleton to efficiently generate a minimal set of specialized and informative viewpoints for complete coverage. Based on SSD, a hierarchical planner effectively divides the large planning problem into independent sub-problems, enabling parallel planning for each subspace. The carefully designed global and local planning strategies are then incorporated to guarantee both high quality and efficiency in path generation. We conduct extensive benchmark and real-world tests, where FC-Planner computes over 10 times faster compared to state-of-the-art methods with shorter path and more complete coverage. The source code will be made publicly available to benefit the community<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup>. Project page: https://hkust-aerial-robotics.github.io/FC-Planner.

A consistent spatial-temporal coordination across multiple agents is fundamental for collaborative perception, which seeks to improve perception abilities through information exchange among agents. To achieve this spatial-temporal alignment, traditional methods depend on external devices to provide localization and clock signals. However, hardware-generated signals could be vulnerable to noise and potentially malicious attack, jeopardizing the precision of spatial-temporal alignment. Rather than relying on external hardwares, this work proposes a novel approach: aligning by recognizing the inherent geometric patterns within the perceptual data of various agents. Following this spirit, we propose a robust collaborative perception system that operates independently of external localization and clock devices. The key module of our system, FreeAlign, constructs a salient object graph for each agent based on its detected boxes and uses a graph neural network to identify common subgraphs between agents, leading to accurate relative pose and time. We validate FreeAlign on both real-world and simulated datasets. The results show that, the FreeAlign empowered robust collaborative perception system perform comparably to systems relying on precise localization and clock devices. ${\mathbf{Code}}$ will be released.

Visual Place Recognition (VPR) in indoor environments is beneficial to humans and robots for better localization and navigation. It is challenging due to appearance changes at various frequencies, and difficulties of obtaining ground truth metric trajectories for training and evaluation. This paper introduces the NYC-Indoor-VPR dataset, a unique and rich collection of over 36,000 images compiled from 13 distinct crowded scenes in New York City taken under varying lighting conditions with appearance changes. Each scene has multiple revisits across a year. To establish the ground truth for VPR, we propose a semiautomatic annotation approach that computes the positional information of each image. Our method specifically takes pairs of videos as input and yields matched pairs of images along with their estimated relative locations. The accuracy of this matching is refined by human annotators, who utilize our annotation software to correlate the selected keyframes. Finally, we present a benchmark evaluation of several state-of-the-art VPR algorithms using our annotated dataset, revealing its challenge and thus value for VPR research.

Adopting omnidirectional Field of View (FoV) cameras in aerial robots vastly improves perception ability, significantly advancing aerial robotics's capabilities in inspection, reconstruction, and rescue tasks. However, such sensors also elevate system complexity, e.g., hardware design, and corresponding algorithm, which limits researchers from utilizing aerial robots with omnidirectional FoV in their research. To bridge this gap, we propose OmniNxt, a fully open-source aerial robotics platform with omnidirectional perception. We design a high-performance flight controller NxtPX4 and a multi-fisheye camera set for OmniNxt. Meanwhile, the compatible software is carefully devised, which empowers OmniNxt to achieve accurate localization and real-time dense mapping with limited computation resource occupancy. We conducted extensive real-world experiments to validate the superior performance of OmniNxt in practical applications. All the hardware and software are open-access at https://github.com/HKUST-Aerial-Robotics/OmniNxt, and we provide docker images of each crucial module in the proposed system. Project page: https://hkust-aerial-robotics.github.io/OmniNxt.

Object detection and multiple object tracking (MOT) are essential components of self-driving systems. Accurate detection and uncertainty quantification are both critical for onboard modules, such as perception, prediction, and planning, to improve the safety and robustness of autonomous vehicles. Collaborative object detection (COD) has been proposed to improve detection accuracy and reduce uncertainty by leveraging the viewpoints of multiple agents. However, little attention has been paid to how to leverage the uncertainty quantification from COD to enhance MOT performance. In this letter, as the first attempt to address this challenge, we design an uncertainty propagation framework called MOT-CUP. Our framework first quantifies the uncertainty of COD through direct modeling and conformal prediction, and propagates this uncertainty information into the motion prediction and association steps. MOT-CUP is designed to work with different collaborative object detectors and baseline MOT algorithms. We evaluate MOT-CUP on V2X-Sim, a comprehensive collaborative perception dataset, and demonstrate a 2% improvement in accuracy and a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$2.67\times$</tex-math></inline-formula> reduction in uncertainty compared to the baselines, e.g. SORT and ByteTrack. In scenarios characterized by high occlusion levels, our MOT-CUP demonstrates a noteworthy 4.01% improvement in accuracy. MOT-CUP demonstrates the importance of uncertainty quantification in both COD and MOT, and provides the first attempt to improve the accuracy and reduce the uncertainty in MOT based on COD through uncertainty propagation.

In recent years, deep learning techniques have shown significant potential for improving video quality assessment (VQA), achieving higher correlation with subjective opinions compared to conventional approaches. However, the development of deep VQA methods has been constrained by the limited availability of large-scale training databases and ineffective training methodologies. As a result, it is difficult for deep VQA approaches to achieve consistently superior performance and model generalization. In this context, this paper proposes new VQA methods based on a two-stage training methodology which motivates us to develop a large-scale VQA training database without employing human subjects to provide ground truth labels. This method was used to train a new transformer-based network architecture, exploiting quality ranking of different distorted sequences rather than minimizing the difference from the ground-truth quality labels. The resulting deep VQA methods (for both full reference and no reference scenarios), FR- and NR-RankDVQA, exhibit consistently higher correlation with perceptual quality compared to the state-of-the-art conventional and deep VQA methods, with average SROCC values of 0.8972 (FR) and 0.7791 (NR) over eight test sets without performing cross-validation. The source code of the proposed quality metrics and the large training database are available at https://chenfeng-bristol.github.io/RankDVQA.

<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Goal:</i> Distance information is highly requested in assistive smartphone Apps by people who are blind or low vision (PBLV). However, current techniques have not been evaluated systematically for accuracy and usability. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Methods:</i> We tested five smartphone-based distance-estimation approaches in the image center and periphery at 1-3 meters, including machine learning (CoreML), infrared grid distortion (IR_self), light detection and ranging (LiDAR_back), and augmented reality room-tracking on the front (ARKit_self) and back-facing cameras (ARKit_back). <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Results:</i> For accuracy in the image center, all approaches had <±2.5cm average error, except CoreML which had ±5.2-6.2cm average error at 2-3 meters. In the periphery, all approaches were more inaccurate, with CoreML and IR_self having the highest average errors at ±41cm and ±32cm respectively. For usability, CoreML fared favorably with the lowest central processing unit usage, second lowest battery usage, highest field-of-view, and no specialized sensor requirements. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Conclusions:</i> We provide key information that helps design reliable smartphone-based visual assistive technologies to enhance the functionality of PBLV.

In this study, a group of active peptides named P3 isolated from soybean protein is used to modify the surface of titanium sheets. The modified titanium sheets are characterized by X-ray photoelectron spectroscopy and contact angle measurement. The effects of modified titanium sheets coated with P3 on the adhesion, proliferation, differentiation and mineralization of MC3T3-E1 cells are investigated. Amino acid analysis showed that the hydrophobic amino acids and aromatic amino acids are enriched in P3. After being coated with P3, the surface nitrogen content of titanium sheets is significantly increased in a dose-dependent manner. The contact angle and surface hydrophobicity of titanium sheets are significantly changed (p &lt; 0.05). In addition, the adhesion and proliferation activity of MC3T3-E1 cells on titanium sheets are significantly enhanced by P3 coating (p &lt; 0.05). Further studies indicated that P3 coating increased the expression of alkaline phosphatase, osteocalcin, consequently promoting the differentiation and mineralization of MC3T3-E1 cells on the surface of titanium sheets. In conclusion, P3 coating enhances the surface hydrophobicity, leading to a significant increase in the cell adhesion, consequently promoting the differentiation and mineralization of MC3T3-E1 cells on the surface titanium sheets, suggesting that P3 may serve as a coating material for titanium-based implants to facilitate fracture healing and other applications.

The purpose of this study is to explore the difference of the influence of motor imagery with different speeds on the maximum power of the exercisers, to find the image mode suitable for improving the performance of maximum power, and to explore the possible functional differences between motor imagery with different speeds. A mixed study design of 4 groups×2 tests was adopted. 79 subjects were randomly divided into slow group (n=19), real speed group (n=19), fast group (n=22) and a control group (n=19). The experiment is divided into three parts: pretest of elbow bending strength, image training and post-test of elbow bending strength. The results showed that the strength performance of the real speed group and the fast group increased significantly after the intervention (p<0.01), while there was no significant difference between the slow speed group and the control group. Comparing the changes in the maximum strength of elbow flexion measured before and after the members of different speed groups, it was found that the changes between groups were significantly different (p<0.01). Specifically, the growth of the maximum strength of elbow flexion in the fast group (p<0.01) and the real speed group (p=0.01) was significantly greater than that in the control group, while there was no significant difference in the growth of elbow flexion strength between the other groups. Conclusion: (1) Real speed and fast motor image training can promote the strength performance of the subjects; (2) When using different motor imagery training speeds, subjects have different functional understandings of the content of imagery.

The aim of this study is to explore and evaluate the effect of dynamic coupling compensation on fuzzy control in cold storage refrigeration systems. By analysing the complexity of the cold storage refrigeration system and the coupling relationship between the components in it, a fuzzy control strategy integrating dynamic coupling compensation is proposed. The control strategy aims to regulate the dynamic characteristics of the refrigeration system through coupling compensation to improve the stability and control performance of the system. In the study, firstly, the coupling relationship of each component in the cold storage refrigeration system is described, and the theoretical basis of dynamic coupling compensation control is elaborated. Secondly, a controller model with dynamic coupling compensation is established based on fuzzy control theory. The model combines the advantages of fuzzy logic and neural network control to adapt to the characteristics of the system's dynamic changes and complex coupling relationships. Further, the effectiveness of the proposed control strategy in the cold storage refrigeration system is verified through experimental analyses. The response curves of storage room temperature and superheat under different control strategies were observed and analysed to assess the performance of the controller and the stability of the system. The results show that the fuzzy control strategy with dynamic coupling compensation can significantly improve the stability of the cold storage refrigeration system and effectively reduce the overshoot and fluctuation of the system. However, a certain level of control challenges still emerged in some cases, especially in the regulation of the superheat. These results provide insights into the effectiveness and limitations of the control strategy and provide an important reference for future control system improvement and optimisation.