Spatial commonsense, the knowledge about spatial position and relationship between objects (like the relative size of a lion and a girl, and the position of a boy relative to a bicycle when cycling), is an important part of commonsense knowledge. Although pretrained language models (PLMs) succeed in many NLP tasks, they are shown to be ineffective in spatial commonsense reasoning. Starting from the observation that images are more likely to exhibit spatial commonsense than texts, we explore whether models with visual signals learn more spatial commonsense than text-based PLMs. We propose a spatial commonsense benchmark that focuses on the relative scales of objects, and the positional relationship between people and objects under different actions. We probe PLMs and models with visual signals, including vision-language pretrained models and image synthesis models, on this benchmark, and find that image synthesis models are more capable of learning accurate and consistent spatial knowledge than other models. The spatial knowledge from image synthesis models also helps in natural language understanding tasks that require spatial commonsense.
In this paper we present a new thermographic image database suitable for the analysis of automatic focus measures. This database consists of 8 different sets of scenes, where each scene contains one image for 96 different focus positions. Using this database we evaluate the usefulness of six focus measures with the goal to determine the optimal focus position. Experimental results reveal that an accurate automatic detection of optimal focus position is possible, even with a low computational burden. We also present an acquisition tool able to help the acquisition of thermal images. To the best of our knowledge, this is the first study about automatic focus of thermal images.
In this paper we consider the localization problem for a visual sensor network. Inspired by the alternate attitude and position distributed optimization framework discussed in [1], we propose an estimation scheme that exploits the unit dual quaternion algebra to describe the sensors pose. This representation is beneficial in the formulation of the optimization scheme allowing to solve the localization problem without designing two interlaced position and orientation estimators, thus improving the estimation error distribution over the two pose components and the overall localization performance. Furthermore, the numerical experimentation asserts the robustness of the proposed algorithm w.r.t. the initial conditions.
Data science and machine learning provide indispensable techniques for understanding phenomena at scale, but the discretionary choices made when doing this work are often not recognized. Drawing from qualitative research practices, we describe how the concepts of positionality and reflexivity can be adapted to provide a framework for understanding, discussing, and disclosing the discretionary choices and subjectivity inherent to data science work. We first introduce the concepts of model positionality and computational reflexivity that can help data scientists to reflect on and communicate the social and cultural context of a model's development and use, the data annotators and their annotations, and the data scientists themselves. We then describe the unique challenges of adapting these concepts for data science work and offer annotator fingerprinting and position mining as promising solutions. Finally, we demonstrate these techniques in a case study of the development of classifiers for toxic commenting in online communities.
Recently, various multimodal networks for Visually-Rich Document Understanding(VRDU) have been proposed, showing the promotion of transformers by integrating visual and layout information with the text embeddings. However, most existing approaches utilize the position embeddings to incorporate the sequence information, neglecting the noisy improper reading order obtained by OCR tools. In this paper, we propose a robust layout-aware multimodal network named XYLayoutLM to capture and leverage rich layout information from proper reading orders produced by our Augmented XY Cut. Moreover, a Dilated Conditional Position Encoding module is proposed to deal with the input sequence of variable lengths, and it additionally extracts local layout information from both textual and visual modalities while generating position embeddings. Experiment results show that our XYLayoutLM achieves competitive results on document understanding tasks.
Pressure ulcers are a challenge for patients and healthcare professionals. In the UK, 700,000 people are affected by pressure ulcers each year. Treating them costs the National Health Service {\pounds}3.8 million every day. Their etiology is complex and multifactorial. However, evidence has shown a strong link between old age, disease-related sedentary lifestyles and unhealthy eating habits. Pressure ulcers are caused by direct skin contact with a bed or chair without frequent position changes. Urinary and faecal incontinence, diabetes, and injuries that restrict body position and nutrition are also known risk factors. Guidelines and treatments exist but their implementation and success vary across different healthcare settings. This is primarily because healthcare practitioners have a) minimal experience in dealing with pressure ulcers, and b) a general lack of understanding of pressure ulcer treatments. Poorly managed, pressure ulcers lead to severe pain, poor quality of life, and significant healthcare costs. In this paper, we report the findings of a clinical trial conducted by Mersey Care NHS Foundation Trust that evaluated the performance of a faster region-based convolutional neural network and mobile platform that categorised and documented pressure ulcers. The neural network classifies category I, II, III, and IV pressure ulcers, deep tissue injuries, and unstageable pressure ulcers. Photographs of pressure ulcers taken by district nurses are transmitted over 4/5G communications to an inferencing server for classification. Classified images are stored and reviewed to assess the model's predictions and relevance as a tool for clinical decision making and standardised reporting. The results from the study generated a mean average Precision=0.6796, Recall=0.6997, F1-Score=0.6786 with 45 false positives using an @.75 confidence score threshold.
In this paper, we develop position embedding transformation (PETR) for multi-view 3D object detection. PETR encodes the position information of 3D coordinates into image features, producing the 3D position-aware features. Object query can perceive the 3D position-aware features and perform end-to-end object detection. PETR achieves state-of-the-art performance (50.4% NDS and 44.1% mAP) on standard nuScenes dataset and ranks 1st place on the benchmark. It can serve as a simple yet strong baseline for future research. Code is available at \url{https://github.com/megvii-research/PETR}.
Cross-view image geolocalization provides an estimate of an agent's global position by matching a local ground image to an overhead satellite image without the need for GPS. It is challenging to reliably match a ground image to the correct satellite image since the images have significant viewpoint differences. Existing works have demonstrated localization in constrained scenarios over small areas but have not demonstrated wider-scale localization. Our approach, called Wide-Area Geolocalization (WAG), combines a neural network with a particle filter to achieve global position estimates for agents moving in GPS-denied environments, scaling efficiently to city-scale regions. WAG introduces a trinomial loss function for a Siamese network to robustly match non-centered image pairs and thus enables the generation of a smaller satellite image database by coarsely discretizing the search area. A modified particle filter weighting scheme is also presented to improve localization accuracy and convergence. Taken together, WAG's network training and particle filter weighting approach achieves city-scale position estimation accuracies on the order of 20 meters, a 98% reduction compared to a baseline training and weighting approach. Applied to a smaller-scale testing area, WAG reduces the final position estimation error by 64% compared to a state-of-the-art baseline from the literature. WAG's search space discretization additionally significantly reduces storage and processing requirements.
Estimating human vital signs in a contactless non-invasive method using radar provides a convenient method in the medical field to conduct several health checkups easily and quickly. In addition to monitoring while sitting and sleeping, the standing position has aroused interest for both the industrial and medical fields. However, it is more challenging due to the micro motions induced by the body for balancing that may cause false respiratory rate estimation. In this work, we focus on the measurement of the respiratory rate of a standing person accurately with the capability of heavy breath detection and estimation. Multiple estimation approaches are presented and compared, including spectral estimation, deep-learning-based approaches, and adaptive peak selection with Kalman filtering. The latest technique is showing the best performance with an absolute error rate of 1.5 bpm, when compared to a Vernier Go Direct\textsuperscript{\textregistered} respiration belt.
We solve object localisation in partial scenes, a new problem of estimating the unknown position of an object (e.g. where is the bag?) given a partial 3D scan of a scene. The proposed solution is based on a novel scene graph model, the Spatial Commonsense Graph (SCG), where objects are the nodes and edges define pairwise distances between them, enriched by concept nodes and relationships from a commonsense knowledge base. This allows SCG to better generalise its spatial inference over unknown 3D scenes. The SCG is used to estimate the unknown position of the target object in two steps: first, we feed the SCG into a novel Proximity Prediction Network, a graph neural network that uses attention to perform distance prediction between the node representing the target object and the nodes representing the observed objects in the SCG; second, we propose a Localisation Module based on circular intersection to estimate the object position using all the predicted pairwise distances in order to be independent of any reference system. We create a new dataset of partially reconstructed scenes to benchmark our method and baselines for object localisation in partial scenes, where our proposed method achieves the best localisation performance.
In this paper, we propose a weakly-supervised approach for 3D object detection, which makes it possible to train a strong 3D detector with position-level annotations (i.e. annotations of object centers). In order to remedy the information loss from box annotations to centers, our method, namely Back to Reality (BR), makes use of synthetic 3D shapes to convert the weak labels into fully-annotated virtual scenes as stronger supervision, and in turn utilizes the perfect virtual labels to complement and refine the real labels. Specifically, we first assemble 3D shapes into physically reasonable virtual scenes according to the coarse scene layout extracted from position-level annotations. Then we go back to reality by applying a virtual-to-real domain adaptation method, which refine the weak labels and additionally supervise the training of detector with the virtual scenes. Furthermore, we propose a more challenging benckmark for indoor 3D object detection with more diversity in object sizes to better show the potential of BR. With less than 5% of the labeling labor, we achieve comparable detection performance with some popular fully-supervised approaches on the widely used ScanNet dataset. Code is available at: https://github.com/wyf-ACCEPT/BackToReality
Reinforcement learning (RL) has become a promising approach to developing controllers for quadrupedal robots. Conventionally, an RL design for locomotion follows a position-based paradigm, wherein an RL policy outputs target joint positions at a low frequency that are then tracked by a high-frequency proportional-derivative (PD) controller to produce joint torques. In contrast, for the model-based control of quadrupedal locomotion, there has been a paradigm shift from position-based control to torque-based control. In light of the recent advances in model-based control, we explore an alternative to the position-based RL paradigm, by introducing a torque-based RL framework, where an RL policy directly predicts joint torques at a high frequency, thus circumventing the use of a PD controller. The proposed learning torque control framework is validated with extensive experiments, in which a quadruped is capable of traversing various terrain and resisting external disturbances while following user-specified commands. Furthermore, compared to learning position control, learning torque control demonstrates the potential to achieve a higher reward and is more robust to significant external disturbances. To our knowledge, this is the first sim-to-real attempt for end-to-end learning torque control of quadrupedal locomotion.
This work addresses the problem of recognizing action categories in videos when no training examples are available. The current state-of-the-art enables such a zero-shot recognition by learning universal mappings from videos to a semantic space, either trained on large-scale seen actions or on objects. While effective, we find that universal action and object mappings are biased to specific regions in the semantic space. These biases lead to a fundamental problem: many unseen action categories are simply never inferred during testing. For example on UCF-101, a quarter of the unseen actions are out of reach with a state-of-the-art universal action model. To that end, this paper introduces universal prototype transport for zero-shot action recognition. The main idea is to re-position the semantic prototypes of unseen actions by matching them to the distribution of all test videos. For universal action models, we propose to match distributions through a hyperspherical optimal transport from unseen action prototypes to the set of all projected test videos. The resulting transport couplings in turn determine the target prototype for each unseen action. Rather than directly using the target prototype as final result, we re-position unseen action prototypes along the geodesic spanned by the original and target prototypes as a form of semantic regularization. For universal object models, we outline a variant that defines target prototypes based on an optimal transport between unseen action prototypes and object prototypes. Empirically, we show that universal prototype transport diminishes the biased selection of unseen action prototypes and boosts both universal action and object models for zero-shot classification and spatio-temporal localization.
Recently, vision transformers started to show impressive results which outperform large convolution based models significantly. However, in the area of small models for mobile or resource constrained devices, ConvNet still has its own advantages in both performance and model complexity. We propose ParC-Net, a pure ConvNet based backbone model that further strengthens these advantages by fusing the merits of vision transformers into ConvNets. Specifically, we propose position aware circular convolution (ParC), a light-weight convolution op which boasts a global receptive field while producing location sensitive features as in local convolutions. We combine the ParCs and squeeze-exictation ops to form a meta-former like model block, which further has the attention mechanism like transformers. The aforementioned block can be used in plug-and-play manner to replace relevant blocks in ConvNets or transformers. Experiment results show that the proposed ParC-Net achieves better performance than popular light-weight ConvNets and vision transformer based models in common vision tasks and datasets, while having fewer parameters and faster inference speed. For classification on ImageNet-1k, ParC-Net achieves 78.6% top-1 accuracy with about 5.0 million parameters, saving 11% parameters and 13% computational cost but gaining 0.2% higher accuracy and 23% faster inference speed (on ARM based Rockchip RK3288) compared with MobileViT, and uses only 0.5 times parameters but gaining 2.7% accuracy compared with DeIT. On MS-COCO object detection and PASCAL VOC segmentation tasks, ParC-Net also shows better performance. Source code is available at https://github.com/hkzhang91/ParC-Net
In recent years, reinforcement learning (RL) has acquired a prominent position in health-related sequential decision-making problems, gaining traction as a valuable tool for delivering adaptive interventions (AIs). However, in part due to a poor synergy between the methodological and the applied communities, its real-life application is still limited and its potential is still to be realized. To address this gap, our work provides the first unified technical survey on RL methods, complemented with case studies, for constructing various types of AIs in healthcare. In particular, using the common methodological umbrella of RL, we bridge two seemingly different AI domains, dynamic treatment regimes and just-in-time adaptive interventions in mobile health, highlighting similarities and differences between them and discussing the implications of using RL. Open problems and considerations for future research directions are outlined. Finally, we leverage our experience in designing case studies in both areas to showcase the significant collaborative opportunities between statistical, RL, and healthcare researchers in advancing AIs.
In this work we present a workflow for designing two linear control techniques applied to the dynamic system quadruple inverted pendulum mounted on a cart (QIP) where the steady state error on cart position is eliminated through a precompensation gain. The first control law designed was based on LQR, technique that stabilizes the system states based on a minimization of the total energy of the system states and the second control technique was pole placement by 2nd order system approximation, this method allowed to define a settling time and overshoot of the desired response of the plant, for this control the system was considered SISO, being the only output of the plant the horizontal position of the cart. To eliminate the steady-state error generated by both linear controllers, an analysis of the stationary response of the system under control in the Laplace Domain was performed. The simulation environment used to validate the results presented in this paper was Simscape of Matlab, it is a simulation environment that allowed to simulate the dynamic of nonlinear mechanical systems. The results obtained indicate that both controllers are able to stabilize the plant and move the cart position to the desired setpoint. Finally, we discuss about the performances of both linear control techniques.
Even though image signals are typically acquired on a regular two dimensional grid, there exist many scenarios where non-regular sampling is possible. Non-regular sampling can remove aliasing. In terms of the non-regular sampling patterns, there is a high degree of freedom in how to actually arrange the sampling positions. In literature, random patterns show higher reconstruction quality compared to regular patterns due to reduced aliasing effects. On the downside, random patterns feature large void areas which is also disadvantageous. In the scope of this work, we present two techniques to design optimized non-regular image sampling patterns for arbitrary sampling densities. Both techniques create incremental sampling patterns, i.e., one pixel position is added in each step until the desired sampling density is reached. Our proposed patterns increase the reconstruction quality by more than +0.5 dB in PSNR for a broad density range. Visual comparisons are provided.
"Human-centered machine learning" (HCML) is a term that describes machine learning that applies to human-focused problems. Although this idea is noteworthy and generates scholarly excitement, scholars and practitioners have struggled to clearly define and implement HCML in computer science. This article proposes practices for human-centered machine learning, an area where studying and designing for social, cultural, and ethical implications are just as important as technical advances in ML. These practices bridge between interdisciplinary perspectives of HCI, AI, and sociotechnical fields, as well as ongoing discourse on this new area. The five practices include ensuring HCML is the appropriate solution space for a problem; conceptualizing problem statements as position statements; moving beyond interaction models to define the human; legitimizing domain contributions; and anticipating sociotechnical failure. I conclude by suggesting how these practices might be implemented in research and practice.
Humanoid robots could replace humans in hazardous situations but most of such situations are equally dangerous for them, which means that they have a high chance of being damaged and falling. We hypothesize that humanoid robots would be mostly used in buildings, which makes them likely to be close to a wall. To avoid a fall, they can therefore lean on the closest wall, as a human would do, provided that they find in a few milliseconds where to put the hand(s). This article introduces a method, called D-Reflex, that learns a neural network that chooses this contact position given the wall orientation, the wall distance, and the posture of the robot. This contact position is then used by a whole-body controller to reach a stable posture. We show that D-Reflex allows a simulated TALOS robot (1.75m, 100kg, 30 degrees of freedom) to avoid more than 75% of the avoidable falls and can work on the real robot.
Multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed. However, the performance of the state-of-the-art models decreases sharply when they are deployed in the real world. We find that the main reason is that real-world applications can only access the text outputs by the automatic speech recognition (ASR) models, which may be with errors because of the limitation of model capacity. Through further analysis of the ASR outputs, we find that in some cases the sentiment words, the key sentiment elements in the textual modality, are recognized as other words, which makes the sentiment of the text change and hurts the performance of multimodal sentiment models directly. To address this problem, we propose the sentiment word aware multimodal refinement model (SWRM), which can dynamically refine the erroneous sentiment words by leveraging multimodal sentiment clues. Specifically, we first use the sentiment word position detection module to obtain the most possible position of the sentiment word in the text and then utilize the multimodal sentiment word refinement module to dynamically refine the sentiment word embeddings. The refined embeddings are taken as the textual inputs of the multimodal feature fusion module to predict the sentiment labels. We conduct extensive experiments on the real-world datasets including MOSI-Speechbrain, MOSI-IBM, and MOSI-iFlytek and the results demonstrate the effectiveness of our model, which surpasses the current state-of-the-art models on three datasets. Furthermore, our approach can be adapted for other multimodal feature fusion models easily. Data and code are available at https://github.com/albertwy/SWRM.