Likelihood estimation of places in local environments

The current study used fMRI in humans to examine goal-directed navigation in an open field environment. We designed a task that required participants to encode survey-level spatial information and subsequently navigate to a goal location in either first person, third person, or survey perspectives. Critically, no distinguishing landmarks or goal location markers were present in the environment, thereby requiring participants to rely on path integration mechanisms for successful navigation. We focused our analysis on mechanisms related to navigation and mechanisms tracking linear distance to the goal location. Successful navigation required translation of encoded survey-level map information for orientation and implementation of a planned route to the goal. Our results demonstrate that successful first and third person navigation trials recruited the anterior hippocampus more than trials when the goal location was not successfully reached. When examining only successful trials, the retrosplenial and posterior parietal cortices were recruited for goal-directed navigation in both first person and third person perspectives. Unique to first person perspective navigation, the hippocampus was recruited to path integrate self-motion cues with location computations toward the goal location. Last, our results demonstrate that the hippocampus supports goal-directed navigation by actively tracking proximity to the goal throughout navigation. When using path integration mechanisms in first person and third person perspective navigation, the posterior hippocampus was more strongly recruited as participants approach the goal. These findings provide critical insight into the neural mechanisms by which we are able to use map-level representations of our environment to reach our navigational goals.

Place recognition is a challenging task in computer vision, crucial for enabling autonomous vehicles and robots to navigate previously visited environments. While significant progress has been made in learnable multimodal methods that combine onboard camera images and LiDAR point clouds, the full potential of these methods remains largely unexplored in localization applications. In this article, we study the impact of leveraging a multi-camera setup and integrating diverse data sources for multimodal place recognition, incorporating explicit visual semantics and text descriptions. Our proposed neural network-based method named MSSPlace utilizes images from multiple cameras, LiDAR point clouds, semantic segmentation masks, and text annotations to generate comprehensive place descriptors. We employ a late fusion approach to integrate these modalities, providing a unified representation. Through extensive experiments on the Oxford RobotCar and NCLT datasets, we systematically analyze the impact of each data source on the overall quality of place descriptors. Our experiments demonstrate that combining data from multiple sensors significantly improves place recognition model performance compared to single modality approaches and leads to state-of-the-art quality. We also show that separate usage of visual or textual semantics (which are more compact representations of sensory data) can achieve promising results in place recognition. The code for our method is publicly available: <uri>https://github.com/alexmelekhin/MSSPlace</uri>

Vision-based place recognition in underwater environments is a key component for autonomous robotic exploration. However, this task can be very challenging due to the inherent properties of this kind of places such as: color distortion, poor visibility, perceptual aliasing and dynamic illumination. In this paper, we present a method for vision-based place recognition in coral reefs. Our method relies on using the Bag-of-Words (BoW) approach to derive a descriptor, for the whole image, from a set of relevant regions, which are extracted by utilizing a visual attention algorithm. The descriptor for each relevant region is built by using an histogram of the chromatic channels of the CIE-Lab color space. We present results of our method for a place recognition task in real life videos as well as comparisons of our method against other popular techniques. It can be seen that our approach performs better in most of the cases.

Visual place recognition is a challenging task for applications such as autonomous driving navigation and mobile robot localization. Distracting elements presenting in complex scenes often lead to deviations in the perception of visual place. To address this problem, it is crucial to integrate information from only task-relevant regions into image representations. In this paper, we introduce a novel holistic place recognition model, TransVPR, based on vision Transformers. It benefits from the desirable property of the self-attention operation in Transformers which can naturally aggregate task-relevant features. Attentions from multiple levels of the Transformer, which focus on different regions of interest, are further combined to generate a global image representation. In addition, the output tokens from Transformer layers filtered by the fused attention mask are considered as key-patch descriptors, which are used to perform spatial matching to re-rank the candidates retrieved by the global image features. The whole model allows end-to-end training with a single objective and image-level supervision. TransVPR achieves state-of-the-art performance on several real-world benchmarks while maintaining low computational time and storage requirements.

We propose an end-to-end place recognition model based on a novel deep neural network. First, we propose to exploit the spatial pyramid structure of the images to enhance the vector of locally aggregated descriptors (VLAD) such that the enhanced VLAD features can reflect the structural information of the images. To encode this feature extraction into the deep learning method, we build a spatial pyramid-enhanced VLAD (SPE-VLAD) layer. Next, we impose weight constraints on the terms of the traditional triplet loss (T-loss) function such that the weighted T-loss (WT-loss) function avoids the suboptimal convergence of the learning process. The loss function can work well under weakly supervised scenarios in that it determines the semantically positive and negative samples of each query through not only the GPS tags but also the Euclidean distance between the image representations. The SPE-VLAD layer and the WT-loss layer are integrated with the VGG-16 network or ResNet-18 network to form a novel end-to-end deep neural network that can be easily trained via the standard backpropagation method. We conduct experiments on three benchmark data sets, and the results demonstrate that the proposed model defeats the state-of-the-art deep learning approaches applied to place recognition.

The Wesenheit function is widely used to reduce the effects of interstellar reddening in distance measurements. Its construction, however, relies on the assumption of a universal extinction curve and on fixed values of the total-to-selective extinction ratio, R V . Recent studies have shown that R V varies significantly across the Milky Way and between different galaxies, raising concerns about systematic biases in Wesenheit magnitudes and period–Wesenheit relations. In this work, we discuss the impact of nonuniversal extinction on Wesenheit indices by combining the R V -dependent extinction curve with a grid of stellar atmosphere models. We compute the integrated extinction in optical and near-infrared passbands, derive R V -dependent R coefficients for multiple Wesenheit indices, and examine how changes in R V propagate into Wesenheit magnitudes and Cepheid distances in our Galaxy. We find that the R coefficients in the Wesenheit functions vary strongly with R V . For classical Cepheids in the Milky Way disk, variations of R V within the typical observed range (2.6–3.6) can lead to substantial differences in the Wesenheit function, reaching ± 0.7 mag from the mean for the Gaia-based Wesenheit index W G and resulting in distance errors of almost 40%. Near-infrared Wesenheit indices are much less sensitive to R V changes. Our results clearly show that accounting for variable R V is essential when applying period–Wesenheit relations, particularly in the optical regime, or that near- or mid-infrared-based distances should be used. While we present this effect for classical Cepheids, it applies to all pulsating stars for which period–Wesenheit relations are used to infer distances.

The aim of the present experiment was to investigate the extent to which the orientation of objects drawn in side view is a function of culture, age, and manual dominance. Right-handed French and Syrian children aged between 6 and 10 years and right-handed adults from the same cultural groups were asked to produce side views of faces, vehicles (car and airplane), self-centered tools with a handle (mug and toothbrush), object-centered tools with a handle (jug and hammer), and animals (dog and fish) using their dominant and nondominant hands. The French participants exhibited a leftward directional bias, whereas their Syrian counterparts displayed a rightward bias. However, no differences between the two cultural groups were observed in the 6-year-olds, who did not present any systematic directional bias in their drawings. Furthermore, regardless of culture, the children did not modify the orientation of the objects in their drawings as a function of the hand used, whereas the hand effect was strong in adults. Finally, despite their directional bias toward the right, the Syrian participants tended to draw the self- and object-centered tools facing leftward. These results are discussed with reference to the current literature on this topic.

This paper argues that a focus on recognition helps in analysing a series of underlying- and salient- yet often under-studied aspects in the process of European integration. The first section revisits the debates in IR literature on recognition and introduces a complementary analytical framework developed around a functionalist typology of recognition in IR. This builds, first, on three models of recognition, that is, a struggle for ethical community (Hegel); a struggle for justice (Honneth) and a struggle for identity (Ringmar). Second, it involves three structuring mechanisms, that is, institutional biases; stigmatisation; and pecking orders (status hierarchies). The third section applies this framework and focuses on the dynamics of recognition struggles revolving around the notions of ethical community, justice and identity in the EU. The analysis operates on two levels: first, on the level of individual member states, it offers an illustrative case study of relations between Viktor Orbán’s government in Hungary and the EU’s institutions and governments of fellow member states. Second, on the polity-level, it assesses a number of recent crises in the EU in the last two decades and the shifting patterns of in-group/out-group recognition dynamics in the EU.

The Finapres non-invasive blood pressure (BP) monitor uses the method of Penaz to indirectly record the arterial waveform; studies on its accuracy have suggested little systematic bias vs intra-arterial pressure (IAP) but substantial variability. Inconsistency between studies, in respect of the magnitude, direction and variance of bias, was described in the validation studies against the direct IAP. We have employed a novel re-sampling statistical method to combine the data from 20 published studies; a robust overall estimate of the accuracy and precision of the Finapres was thereby obtained. Based on 449 patients and 4490 re-samples, the average Finapres systolic bias (IAP- Finapres) was 2.2 mm Hg (s.d.+/-12.4) with limits of agreement (bias+/-2 s.d.) of -22.6 and 26.9 mm Hg. The average precision was 12.1 mm Hg (s.d.+/-8.4). The Finapres diastolic bias was -0.3 mm Hg (s.d.+/-7.9) with the limits of agreement of -16.1 and 15.5 mm Hg. The average precision was 7.6 mm Hg (s.d.+/-5.3). The average Finapres mean arterial pressure bias was 2.1 mm Hg (s.d.+/-8.6) with precision of 7.6 mm Hg (s.d.+/-5.3). The calculated percentage of Finapres systolic values expected to fall within +/-5 or +/-10 of the direct intra-arterial pressure was 35.9% and 73.1%, respectively. The calculated precision of the Finapres systolic pressure between 0-5 mm Hg was 1.6% and between 0-10 mm Hg 36.4%. The comparable values for Finapres diastolic BP for accuracy were 63.5% and 92.8% and for precision 23.1% and 79.2%. The Finapres device can provide an accurate estimate of diastolic and mean arterial pressure compared with the intra-arterial record; the apparent inaccuracy of the Finapres systolic pressure may have a physiological explanation. When the Finapres device is used in experimental or in clinical situations, then calibration against a reliable reference arterial pressure is desirable to obviate the possibility of an 'offset' error.

Electrophysiological experiments have shown that during voluntary movement, the activity of kinesthetic receptors relevant to actively contracting muscles is modified. Furthermore, efferent motor commands interact with the central transmission and processing of kinesthetic sensory input. This article addresses itself to the question of whether the efferent motor commands of voluntary movement influence the perception and processing of kinesthetic spatial information. In two experiments, subjects moved around unfamiliar and unseen criterion stencil patterns with their right hands. In the first experiment subjects were required to concurrently track the constrained movements with their left hands. In the second experiment, following movement around a criterion pattern, the subjects had to recall it with their right hands. Tracking and recall movements in the two experiments showed a systematic spatial bias in the direction in which the right-hand constrained movements were commanded. It is argued that when a limb movement is contrained, the interaction between the efferent command pattern and the returning kinesthetic sensory input results in its position being perceived with a constant error in the direction in which the movement was commanded.

We use simulated type Ia supernova (SN Ia) samples, including both photometry and spectra, to perform the first direct validation of cosmology analysis using the SALT-II light curve model. This validation includes residuals from the light curve training process, systematic biases in SN Ia distance measurements, and a bias on the dark energy equation of state parameter w. Using the SN-analysis package SNANA, we simulate and analyze realistic samples corresponding to the data samples used in the SNLS3 analysis: ~120 low-redshift (z &lt; 0.1) SNe Ia, ~255 Sloan Digital Sky Survey SNe Ia (z &lt; 0.4), and ~290 SNLS SNe Ia (z ≤ 1). To probe systematic uncertainties in detail, we vary the input spectral model, the model of intrinsic scatter, and the smoothing (i.e., regularization) parameters used during the SALT-II model training. Using realistic intrinsic scatter models results in a slight bias in the ultraviolet portion of the trained SALT-II model, and w biases (w (input) – w (recovered)) ranging from –0.005 ± 0.012 to –0.024 ± 0.010. These biases are indistinguishable from each other within the uncertainty, the average bias on w is –0.014 ± 0.007.

The ability of normal hearing listeners to spatially localize single-syllable speech sounds was examined via a two-interval, two alternative forced choice (2AFC) spatial offset discrimination task. To examine directional biases in horizontal sound localization, this task required listeners to observe the direction of an unseen second talker’s voice immediately after hearing a first talker’s voice in an alternative direction. Listeners reported only the relative frontward versus rearward spatial offset between the two voices heard during each trial. Headphone-based virtual acoustic simulation enabled blind listening tests that were completed without feedback to the listener, revealing the ease with which listeners judged the relative spatial offset of the second speech sound, although those judgments were strongly biased by differences in the source spectra. The single-syllable speech-sound stimuli were processed using individually measured head-related transfer functions (HRTFs) of 15 listeners for whom individually measured earphone transfer functions (ETFs) were also employed. While analysis of the 2AFC response data via receiver operating characteristic (ROC) curves showed a generally high level of directional sensitivity, systematic biases were observed that could be explained via the differences in vowel coloration between the first and second source presented during each trial.

Land surface temperature (LST) is fundamental for monitoring surface energy balance and environmental dynamics, with remote sensing providing the primary means of acquisition. However, directional anisotropy (DA) introduces systematic bias in satellite-derived LST products, particularly over complex landscapes. This study examines the impact of angular effects on LST retrievals from three leading East Asian geostationary satellites (FengYun 4A, FengYun 4B, and Himawari 9) across Hunan Province, China, using integrated comparison with in situ measurements and reanalysis data. Results show that all products exhibit a systematic cold bias, with FY4B achieving the highest accuracy. Diurnal retrieval precision increases with higher solar zenith angles (SZA), while no consistent relationship is observed between viewing zenith angle (VZA) and retrieval accuracy. Notably, the retrieval bias of the FY4 series increases significantly when the sun and sensor are aligned in azimuth, particularly when the relative azimuth angle (RAA) is less than or equal to 30°. Parametric modeling reveals that emissivity kernel-induced anisotropy is the principal driver of significant LST deviations in central Hunan, while solar kernel effects result in LST overestimation in mountainous regions and underestimation in plains. Increases in elevation or vegetation density reduce emissivity-induced errors but amplify errors caused by shadowing and sunlit effects. Emissivity anisotropy is thus identified as the primary source of LST DA. These findings deepen the understanding of LST DA in remote sensing and provide essential guidance for refining retrieval algorithms and improving the applicability of LST products in complex terrains.

With recent advancements in speech recognition, it is crucial to ensure that automatic speech recognition (ASR) systems do not exhibit systematic biases, such as those related to gender, age, accent, and dialect. Although research has extensively examined systematic biases such as those related to gender, age, accent, and dialect, for high-resource languages, research on low-resource African languages remains limited. This systematic literature review synthesizes evidence on bias evaluation and mitigation in ASR models for African languages, adhering to the PRISMA reporting guidelines. Our analysis reveals that most biases stem from data imbalances and limited linguistic diversity in training datasets, resulting in disproportionately high error rates for underrepresented speaker groups. Mitigation strategies in African contexts have primarily focused on data-centric methods, including dataset expansion, augmentation, and transfer learning. In contrast, more advanced approaches, including fairness-aware modeling, bias-aware loss functions, adversarial debiasing, and speaker-adaptive techniques, are rarely applied. Gender, accent, and dialect biases dominate the few African studies available, while age and racial biases are almost absent. The limited number of African languages covered highlights the urgent need for more representative and inclusive research. Addressing these gaps will support the development of fairer and more robust ASR technologies across the continent.

We propose a novel algorithm for improved localization of the mobile robots by fusing RFID localization system and ultra-sonic sensors. Because there are uncertainties in previous RFID localization system, we are focusing on the sensor fusion system, which removes the uncertainty of the robot location using the distance measurements by ultra-sonic sensors. We divide into 2 parts, which are GPE (global position estimation) by RFID localization system and LEC (local environment cognition) by ultra-sonic sensor. The scheme to remove the uncertainty is proposed, which are the modeling about the measurement noise for RFID localization system and to estimate the straight obstacle and circular obstacle in well-structural indoor environment by the ultra-sonic sensors. The hierarchical localization algorithm is proposed to estimate the position of the mobile robot from GPE and LEC and for robust localization under the unknown indoor environment.