Object Size Research Articles

What Do Self-Supervised Speech Models Know About Words?

Abstract Many self-supervised speech models (S3Ms) have been introduced over the last few years, improving performance and data efficiency on various speech tasks. However, these empirical successes alone do not give a complete picture of what is learned during pre-training. Recent work has begun analyzing how S3Ms encode certain properties, such as phonetic and speaker information, but we still lack a proper understanding of knowledge encoded at the word level and beyond. In this work, we use lightweight analysis methods to study segment-level linguistic properties—word identity, boundaries, pronunciation, syntactic features, and semantic features—encoded in S3Ms. We present a comparative study of layer-wise representations from ten S3Ms and find that (i) the frame-level representations within each word segment are not all equally informative, and (ii) the pre-training objective and model size heavily influence the accessibility and distribution of linguistic information across layers. We also find that on several tasks—word discrimination, word segmentation, and semantic sentence similarity—S3Ms trained with visual grounding outperform their speech-only counterparts. Finally, our task-based analyses demonstrate improved performance on word segmentation and acoustic word discrimination while using simpler methods than prior work.1

ML based approach for determining the spatial position and size of objects on images

This paper considers an urgent problem of assessing spatial position and geometric characteristics of environmental objects from images. An approach was developed based on combining the results of object detection using the Mask R-CNN model and the reconstruction of depth maps obtained using the RealSense camera. We evaluated the class-averaged values of the relative error in determining the size of objects for test sets of images formed at various levels of scene illumination: 0.1449, 0.3313, 0.6332. Also, within the experiments relative deviation values were obtained when determining the spatial positions of objects: 0.1010, 0.1624, 0.3477. В контексте определения характеристик киберфизической среды одной из актуальных проблем является определение пространственных координат, местоположения и размеров объектов внешнего окружения. В рамках настоящего исследования для решения указанной проблемы был предложен подход к оценке пространственного положения и геометрических характеристик объектов окружающей среды по изображениям, основанный на объединении результатов детектирования объектов с использованием нейросетевой модели Mask R-CNN, а также результатов восстановления карт глубины наблюдаемой сцены с помощью камеры RealSense D435. В работе представлены авторские алгоритмы анализа геометрических свойств наблюдаемых на изображении областей сцены, а также комплементарные алгоритмы оценки размеров и положения объектов. Оценка качества разработанного подхода проводилась на основе тестового набора данных, включающего 12 000 изображений, сформированных при трех различных уровнях освещенности сцены: 100, 70 и 50 %. В рамках эксперимента найдены усредненные по классам значения относительной ошибки вычисления размеров объектов для соответствующих наборов изображений: 0.1449, 0.3313, 0.6332. Значения относительного отклонения при определении пространственного положения объектов по результатам тестирования составили 0.1010, 0.1624, 0.3477. Таким образом, предлагаемое решение позволяет успешно определять размеры и пространственное положение объектов различных классов по изображениям сцены с уровнями освещенности 100 и 70 %.

C3TB-YOLOv5: integrated YOLOv5 with transformer for object detection in high-resolution remote sensing images

ABSTRACT In the realm of object detection from high-resolution remote sensing images (HRRSIs), the existing YOLOv5 methods encounter several challenges, including dense object arrangements, small object sizes, and complex backgrounds. To tackle these challenges, we propose a novel approach called C3TB-YOLOv5, which combines traditional YOLOv5 with the Transformer model to detect objects in HRRSIs. Unlike conventional YOLOv5 methods that primarily focus on capturing local information from remote sensing scenes, our C3TB-YOLOv5 method incorporates global information through the introduction of a new C3TB module. This module, based on the Transformer multi-head attention mechanism (AM), consists of two branches that extract local and global information from feature maps. By integrating these branches and establishing long-range relationships, our method successfully detects densely arranged small objects in HRRSIs. Furthermore, to improve the accuracy of tiny object detection, a novel detection head has been developed to effectively utilize the unused C3 module, thereby preventing the loss of fine-grained textures and positional features. In addition, we integrate an enhanced SimAM, namely Sim-GMP, into the model to adjust the focus across varying regions, effectively distinguishing the features of interested objects from complex backgrounds. Finally, to address the problem of sample imbalance in remote sensing object detection, the most recent Wise-IoU v3 loss function is employed to improve the accuracy of anchor box predictions for objects. To maintain a high object detection speed, the most critical C3 modules are substituted with the proposed C3TB module for the purpose of striking a good balance between object detection accuracy and model lightweight. Extensive experiments conducted on two remote sensing datasets of NWPU VHR-10 and VisDrone 2019 demonstrates that our method achieves superior object detection performance than state-of-the-art methods.

A miniaturized array microneedle tactile sensor for intelligent object recognition

Tactile sensors can provide intuitive information about the shape, size, and texture of objects, which helps robots make more accurate decisions. In this paper, based on the principle of triboelectrification and electrostatic induction, a miniature tactile sensor (MTS) for smart object recognition is designed using composite materials based on silica gel. Its diameter is only 6 mm, and the electrodes are arranged in a circumferential array, which has excellent shape adaptability and can recognize curved materials with curvature of 1/15 mm−1. By optimizing the sensor structure and process to obtain a microstructure with a spike-like shape, the sensitivity of the sensor is successfully improved, and the smallest and accurately detected surface material side length is 5 mm. In addition, the article provides a smart sensor system that can enhance the recognition function of robots with the help of machine learning technology, and the discrimination accuracy reaches 95.7%, which lays a solid foundation for practical applications.

The visible and thermal light curve of the large Kuiper belt object (50000) Quaoar

Recent stellar occultations have allowed accurate instantaneous size and apparent shape determinations of the large Kuiper belt object (50000) Quaoar and the detection of two rings with spatially variable optical depths. In this paper we present new visible range light curve data of Quaoar from the Kepler/K2 mission, and thermal light curves at 100 and 160 µm obtained with Herschel/PACS. The K2 data provide a single-peaked period of 8.88 h, very close to the previously determined 8.84 h, and it favours an asymmetric double-peaked light curve with a 17.76 h period. We clearly detected a thermal light curve with relative amplitudes of ~ 10% at 100 and at 160 µm. A detailed thermophysical modelling of the system shows that the measurements can be best fit with a triaxial ellipsoid shape, a volume-equivalent diameter of 1090 km, and axis ratios of a/b = 1.19 and b/c = 1.16. This shape matches the published occultation shape, as well as visual and thermal light curve data. The radiometric size uncertainty remains relatively large (±40 km) as the ring and satellite contributions to the system-integrated flux densities are unknown. In the less likely case of negligible ring or satellite contributions, Quaoar would have a size above 1100 km and a thermal inertia ≤ 10 J m−2K−1s−1/2. A large and dark Weywot in combination with a possible ring contribution would lead to a size below 1080 km in combination with a thermal inertia ≳10 J m−2K−1s−1/2, notably higher than that of smaller Kuiper belt objects with similar albedo and colours. We find that Quaoar’s density is in the range 1.67–1.77 g cm−3, significantly lower than previous estimates. This density value closely matches the relationship observed between the size and density of the largest Kuiper belt objects.

Quantitative grain size estimation on airless bodies from the negative polarization branch

Context. Sunlight scattered from the surface of an airless body is generally partially polarized, and the corresponding polarization state includes information about the scattering surface, such as albedo, surface grain sizes, composition, and taxonomic types. Most polarimetric studies on airless bodies thus far have focused on optical wavelengths (λ ≲ 1 µm). Aims. We conducted polarimetry of two large airless bodies, the Dawn mission targets (1) Ceres and (4) Vesta, in the near-infrared region. We further investigated the change in the polarimetric phase curves over the wavelengths expected from previous works. Methods. We used the Nishiharima Infrared Camera (NIC) installed at the Nishi-Harima Astronomical Observatory (NHAO) to observe these objects at multiple geometric configurations in the J, H, and Ks bands (λ ~ 1.2–2.3 µm). Results. Polarimetric parameters were determined and compared with previously reported experimental results. In particular, Vesta exhibits a characteristic change in the negative polarization branch as the wavelength increases to the Ks band, which we interpret as an indication of the dominant existence of D ~ 10–20 µm particles. Our approach is supported by empirical reasoning and coincides well with an independent, theory-driven approach based on thermal modeling. Conclusions. This work demonstrates how near-infrared polarimetry can be utilized to quantitatively determine the particle size of airless objects. This finding will have important implications for asteroid taxonomy and regolith evolution.

Object Depth and Size Estimation Using Stereo-Vision and Integration With SLAM

Object Depth and Size Estimation Using Stereo-Vision and Integration With SLAM

Effects of virtual agents on interaction efficiency and environmental immersion in MR environments

BackgroundPhysical entity interactions in mixed reality (MR) environments aim to harness human capabilities in manipulating physical objects, thereby enhancing virtual environment (VEs) functionality. In MR, a common strategy is to use virtual agents as substitutes for physical entities, balancing interaction efficiency with environmental immersion. However, the impact of virtual agent size and form on interaction performance remains unclear. MethodsTwo experiments were conducted to explore how virtual agent size and form affect interaction performance, immersion, and preference in MR environments. The first experiment assessed five virtual agent sizes (25%, 50%, 75%, 100%, and 125% of physical size). The second experiment tested four types of frames (no frame, consistent frame, half frame, and surrounding frame) across all agent sizes. Participants, utilizing a headmounted display, performed tasks involving moving cups, typing words, and using a mouse. They completed questionnaires assessing aspects such as the virtual environment effects, interaction effects, collision concerns, and preferences. ResultsResults from the first experiment revealed that agents matching physical object size produced the best overall performance. The second experiment demonstrated that consistent framing notably enhances interaction accuracy and speed but reduces immersion. To balance efficiency and immersion, frameless agents matching physical object sizes were deemed optimal. ConclusionsVirtual agents matching physical entity sizes enhance user experience and interaction performance. Conversely, familiar frames from 2D interfaces detrimentally affect interaction and immersion in virtual spaces. This study provides valuable insights for the future development of MR systems.

What drives the automatic retrieval of real-world object size knowledge?

Real-world object size is a behaviorally relevant object property that is automatically retrieved when viewing object images: participants are faster to indicate the bigger of two object images when this object is also bigger in the real world. What drives this size Stroop effect? One possibility is that it reflects the automatic retrieval of real-world size after objects are recognized at the basic level (e.g., recognizing an object as a plane activates large real-world size). An alternative possibility is that the size Stroop effect is driven by automatic associations between low-/mid-level visual features (e.g., rectilinearity) and real-world size, bypassing object recognition. Here, we tested both accounts. In Experiment 1, objects were displayed upright and inverted, slowing down recognition while equating visual features. Inversion strongly reduced the Stroop effect, indicating that object recognition contributed to the Stroop effect. Independently of inversion, however, trial-wise differences in rectilinearity also contributed to the Stroop effect. In Experiment 2, the Stroop effect was compared between manmade objects (for which rectilinearity was associated with size) and animals (no association between rectilinearity and size). The Stroop effect was larger for animals than for manmade objects, indicating that rectilinear feature differences were not necessary for the Stroop effect. Finally, in Experiment 3, unrecognizable "texform" objects that maintained size-related visual feature differences were displayed upright and inverted. Results revealed a small Stroop effect for both upright and inverted conditions. Altogether, these results indicate that the size Stroop effect partly follows object recognition with an additional contribution from visual feature associations. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Grasping tiny objects.

In grasping studies, maximum grip aperture (MGA) is commonly used as an indicator of the object size representation within the visuomotor system. However, a number of additional factors, such as movement safety, comfort, and efficiency, might affect the scaling of MGA with object size and potentially mask perceptual effects on actions. While unimanual grasping has been investigated for a wide range of object sizes, so far very small objects (<5mm) have not been included. Investigating grasping of these tiny objects is particularly interesting because it allows us to evaluate the three most prominent explanatory accounts of grasping (the perception-action model, the digits-in-space hypothesis, and the biomechanical account) by comparing the predictions that they make for these small objects. In the first experiment, participants ( ) grasped and manually estimated the height of square cuboids with heights from 0.5 to 5mm. In the second experiment, a different sample of participants ( ) performed the same tasks with square cuboids with heights from 5 to 20mm. We determined MGAs, manual estimation apertures (MEA), and the corresponding just-noticeable differences (JND). In both experiments, MEAs scaled with object height and adhered to Weber's law. MGAs for grasping scaled with object height in the second experiment but not consistently in the first experiment. JNDs for grasping never scaled with object height. We argue that the digits-in-space hypothesis provides the most plausible account of the data. Furthermore, the findings highlight that the reliability of MGA as an indicator of object size is strongly task-dependent.

Image Quality Enhancement for Digital Breast Tomosynthesis: High-Density Object Artifact Reduction.

Breast cancer has a high incidence and mortality rate among women, early diagnosis is essential as it gives insight regarding the most appropriate therapeutic strategy for each case. Among all imaging diagnostic methods, digital breast tomosynthesis (DBT) is effective for early breast cancer detection. In DBT images, high-density object artifacts are generated when imaging objects with high X-ray absorptivity, which include metal artifacts, ripple artifacts, and deformation artifacts. In this study, we analyze the causes of these artifacts and propose a set of high-density object reconstruction methods based on iterative algorithms. Our method includes a reprojection-based high-density object projection data segmentation algorithm and an iterative reconstruction algorithm based on volume expansion. The experiments on simulation data and the human breast data with artificial surgical needles prove the effectiveness of our method. By using our algorithm, the problem of distorting the shape, size, and position of high-density objects during DBT reconstruction is raised, the influence of these artifacts is reduced, and the quality of the DBT reconstructed image is improved. We hope that our algorithm might contribute to promoting the usage of DBT.

Object-based attention during scene perception elicits boundary contraction in memory.

Boundary contraction and extension are two types of scene transformations that occur in memory. In extension, viewers extrapolate information beyond the edges of the image, whereas in contraction, viewers forget information near the edges. Recent work suggests that image composition influences the direction and magnitude of boundary transformation. We hypothesize that selective attention at encoding is an important driver of boundary transformation effects, selective attention to specific objects at encoding leading to boundary contraction. In this study, one group of participants (N = 36) memorized 15 scenes while searching for targets, while a separate group (N = 36) just memorized the scenes. Both groups then drew the scenes from memory with as much object and spatial detail as they could remember. We asked online workers to provide ratings of boundary transformations in the drawings, as well as how many objects they contained and the precision of remembered object size and location. We found that search condition drawings showed significantly greater boundary contraction than drawings of the same scenes in the memorize condition. Search drawings were significantly more likely to contain target objects, and the likelihood to recall other objects in the scene decreased as a function of their distance from the target. These findings suggest that selective attention to a specific object due to a search task at encoding will lead to significant boundary contraction.

Mean orientation discrimination based on proximal stimuli.

Ensemble perception refers to the ability to accurately and rapidly perceive summary statistical representations of specific features from a group of similar objects. However, the specific type of representation involved in this perception within a three-dimensional (3-D) environment remains unclear. In the context of perspective viewing with stereopsis, distal stimuli can be projected onto the retina as different forms of proximal stimuli based on their distances, despite sharing similar properties, such as object size and spatial frequency. This study aimed to investigate the effects of distal and proximal stimuli on the perception of summary statistical information related to orientation. In our experiment, we presented multiple Gabor patches in a stereoscopic environment, allowing us to measure the discrimination threshold of the mean orientation. The object size and spatial frequency were fixed for all patches regardless of depth. However, the physical angular size and absolute spatial frequency covaried with the depth. The results revealed the threshold elevation with depth expansion, especially when the patches formed two clusters at near and far distances, leading to large variations in their retinotopic representations. This finding indicates a minor contribution of similarity of the distal stimuli. Subsequent experiments demonstrated that the variability in physical angular size of the patches significantly influenced the threshold elevation in contrast to that of binocular disparity and absolute spatial frequency. These findings highlight the critical role of physical angular size variability in perceiving mean orientations within the 3-D space.

Quantifying partial volume effect in SPECT and planar imaging: optimizing region of interest for activity concentration estimation in different sphere sizes.

To quantify the partial volume effect in single photon emission tomography (SPECT) and planar images of Carlson phantom as well as providing an optimum region of interest (ROI) required to more accurately estimate the activity concentration for different sphere sizes. 131 I solution with the 161.16 kBq/ml concentration was uniformly filled into the different spheres of Carlson phantom (cold background condition) with the diameters of 7.3, 9.2, 11.4, 14.3, 17.9, 22.4 and 29.9 mm, and there was no background activity. In the hot background condition, the spheres were filled with the solution of 131 I with the 1276.5 kBq/ml addition to the background activity concentration of 161.16 kBq/ml in all the phantoms. The spheres were mounted inside the phantom and underwent SPECT and planar images. ROI was drawn closely on the boundary of each sphere image and it was extended to extract the true count. In the cold background condition, the recovery coefficient (RC) value for SPECT images ranged between 0.8 and 1.03. However, in planar imaging, the RC value was 0.72 for the smallest sphere size and it increased for larger spheres until 0.98 for 29.9 mm. In the hot background condition, the RC value for sphere diameters larger than 20 mm was overestimated more than in the cold background condition. The ROI/size required to more accurately determine activity concentration for the cold background ranged from 1.18 to 2.7. However, in the hot background condition, this ratio varied from 1.34 to 4.05. In the quantification of partial volume effects, the spill-out effect seems to play a crucial role in the distribution of the image counts beyond the boundaries of the image pixels. However, more investigations are needed to accurately characterize limitations regarding the object size, background levels, and other factors.

Sequence of Simple Digital Technologies for Detection of Platelets in Medical Images

Platelets play an important role in the diagnosis and detection of various diseases, the course of the disease in the future, and the possibility of justifying treatment methods. In this aspect, platelet counting is of key importance. For these purposes, it is important to correctly identify such objects. This served as the basis for the development and generalization of an appropriate medical image analysis procedure. The purpose of this study is to construct a generalized procedure for platelet identification in medical digital images. The work examined at least 30 images containing objects such as platelets. These images are approximately the same type, but with different intensity of the presence of the main objects in the blood smear. Similar but noisy images are also considered. These images were noisy with different types of noise. Thus, a total of at least 120 images were examined. In general, this allows us to evaluate the effectiveness of the proposed procedure for identifying platelets in medical images. This procedure includes simple methods of image analysis such as: binarization, morphological analysis, taking into account the influence of the sizes of different objects and comparative analysis of images at intermediate stages of the study. To summarize the results, estimates such as the percentage of false detection of platelets and the percentage of missed platelets were considered. The platelet identification results that were obtained for non-noisy images are as follows: false platelet isolation was less than 0.1%, missed platelets were within 2-2.5%. The worst result for noisy images is false platelet isolation – within 10% (for images with multiplicative noise), missed platelets – within 7.5-8% (for images with multiplicative noise). It should also be noted that noisy images are characterized by identification of platelets with distortion of their sizes: these sizes are reduced or increased. The percentage of such distortions does not exceed 1.3% (for images with Poisson noise).The problematic aspects of platelet identification in digital medical images are considered. Particular attention is paid to simple methods of digital image processing. Among the problematic aspects of the proposed approach there is a need to clearly take into account the geometric dimensions of platelets. The results obtained are acceptable and can be used as the basis for an automated blood smear analysis system.

Examination of archaeological bronze parts using micro-computed tomography and metallography

Abstract Archeological finds, such as old coins among which some show signs of corrosion, are, in general, unique pieces even though several copies exist of individual types. Subferrati, Roman coins with an iron core and a bronze coating, for instance, could be examined by conventional computed tomography (CT). The CT image allows an easy distinction of corrosion products, metallic iron, and the bronze so that appropriate sections could be selected from the images for metallography. The micro-CT technology even allows examining smaller archaeological objects to get an overview. The resolution of the micro-CT is a function of the object size and lies between 5 and 10 μm. The following metallographic examinations are intended to characterize the microstructure. Apart from these examinations, SEM analyses may also be performed on the sections. Different bronze pieces from cremation and inhumation burials were examined. Micro-CT provides a well recognizable contrast between corrosion products and the metallic bronze. If several similar objects exist, micro-CT examinations allow a decision to be made on which parts are suitable for metallographic analyses and which are not.

Grasping affordance judgments depend on the object emotional value.

The concept of affordance refers to the opportunities for action provided by the environment, often conveyed through visual information. It has been applied to explain visuomotor processing and movement planning. As emotion modulates both visual perception and the motor system, it is reasonable to ask whether emotion can influence affordance judgments. If present, this relationship can have important ontological implications for affordances. Thus, we investigated whether the emotional value of manipulable objects affected the judgment of the appropriate grasping that could be used to interact with them (i.e., their affordance). Volunteers were instructed to use a numerical scale to report their judgment on how an observed object should be grasped. We compared these judgments across emotional categories of objects (pleasant, unpleasant and neutral), while also considering the expected effect of object size. We found that unpleasant objects were rated as more appropriately graspable by a precision grip than pleasant and neutral objects. Simultaneously, smaller object size also favored this judgment. This effect was seen in all emotional categories examined in equal magnitude. Our findings suggest that the emotional value of objects modulates affordance judgments in a way that favors careful manipulation and minimal physical contact with aversive stimuli. Finally, we discuss how this affective aspect of our experience of objects overlaps with what affordances are conceptualized to be, calling for further reexamination of the relationship between affordances and emotions.

Evaluation of the effectiveness of the 5-year rule — impact on the orbital environment at each altitude by reducing the post-mission disposal lifetime

This paper examines the effect of shortening the post-mission disposal (PMD) orbit lifetime (remaining orbit lifetime after PMD) from 25 years to 5 or 1 year. Using NEODEEM, the debris evolutionary model developed by Kyushu University and JAXA, the change in the orbital environment at various altitude bands for long-term stability and short-term safety are discussed. It was confirmed that the PMD compliance rate is more important than shortening the orbit lifetime for improving the long-term stability of the orbital environment. In the short term, the 5- or 1-year rule reduced the collision rate and collision avoidance frequency at an altitude below 700 km, more than the 25-year rule, thus shortening the PMD orbit lifetime improves short-term safety, especially at low altitudes. The paper also evaluates the effect of large constellations (LCs) and it was shown that shortening the PMD orbit lifetime of LCs improves short-term safety, but the long-term impact is dominated by the number of LCs that remain at high operational altitudes due to PMD failure. It shows that the post-mission disposal lifetime and required compliance rate should be set according to the number and size of objects expected to be launched in the future.

FVCNet: Detection obstacle method based on feature visual clustering network in power line inspection

AbstractPower line inspection is an important means to eliminate hidden dangers of power lines. It is a difficult research problem how to solve the low accuracy of power line inspection based on deep neural network (DNN) due to the problems of multi‐view‐shape, small‐size object. In this paper, an automatic detection model based on Feature visual clustering network (FVCNet) for power line inspection is established. First, an unsupervised clustering method for power line inspection is proposed, and applied to construct a detection model which can recognize multi‐view‐shape objects and enhanced object features. Then, the bilinear interpolation method is used to Feature enhancement method, and the enhanced high‐level semantics and low‐level semantics are fused to solve the problems of small object size and single sample. In this paper, FVCNet is applied to the MS‐COCO 2017 data set and self‐made power line inspection data set, and the test accuracy is increased to 61.2% and 82.0%, respectively. Compared with other models, especially for the categories that are greatly affected by multi‐view‐shape, the test accuracy has been improved significantly.

Model-agnostic progressive saliency map generation for object detector

With the widespread adoption of object detection models across various industries, the interpretability of these detectors has become an important research topic. The interpretability of a detector helps humans understand which areas significantly contribute to the model's decision. Furthermore, the interpretability enhances the credibility of detectors and helps identify their strengths and weaknesses. Due to the ability to provide intuitive explanations of models, the saliency map has been widely employed in the field of interpreting deep models. Model-agnostic interpretability methods are more general approaches as they treat the model as a black box without considering its internal complexity structure. However, existing model-agnostic interpretability methods often introduce “noise” into saliency maps by applying random masking and fixed masking granularity. This noise reduces the quality and interpretability of the generated saliency maps. To address this challenge and obtain more interpretable saliency maps for object detection models, this paper proposes a model-agnostic progressive saliency map generation method based on a hierarchical framework called MAPSM. In MAPSM, an adaptive masking partition mechanism is introduced to adapt the masking granularity to different object sizes. Additionally, MAPSM employs a saliency-driven mask generation strategy to effectively reduce the “noise”. Utilizing a hierarchical framework, MAPSM progressively discovers and refines the saliency areas of objects, resulting in more interpretable saliency maps. To evaluate the quality of the saliency maps generated by MAPSM, we compare it with other methods in multiple metrics. Experimental results demonstrate that our method produces saliency maps with better quality and interpretability.