We propose a method to produce continuous stream of novel views under fine-grained (e.g., 1 degree step-size) camera control at interactive rates. A learning pipeline determines the output pixels directly from source color. Injecting geometric transformations, including perspective projection, 3D rotation and translation into network forces implicit reasoning about underlying geometry. The latent geometry representation is compact meaningful transformation, being able geometrically accurate...

In this paper we contribute a simple yet effective approach for estimating 3D poses of multiple people from multi-view images. Our proposed coarse-to-fine pipeline first aggregates noisy 2D observations camera views into space and then associates them individual instances based on confidence-aware majority voting technique. The final pose estimates are attained novel optimization scheme which links high-confidence joint candidates. More-over, statistical parametric body model such as SMPL is...

Handwritten signatures widely exist in our daily lives. The main challenge of signal recognition on handwriting is the development approaches to obtain information effectively. External mechanical signals can be easily detected by triboelectric nanogenerators which provide immediate opportunities for building new types active sensors capable recording handwritten signals. In this work, we report an intelligent human-machine interaction interface based a nanogenerator. Using...

Action detection plays an important role in the field of video understanding and attracts considerable attention last decade. However, current action methods are mainly based on visible videos, few them consider scenes with low-light, where actions difficult to be detected by existing methods, or even human eyes. Compared infrared videos more suitable for dark environment resistant background clutter. In this paper, we investigate temporal problem using which is, best our knowledge, first...

Pose detection of small targets in poor imaging conditions like heavy occlusion and low resolution is still an open challenging task computer vision. For instance, students' poses classrooms that are even indistinguishable to human eyes remains a rather difficult task. Motivated by the success convolutional feature merging locality preserving, authors propose pose framework combining merged region interest (ROI) pooling preserving learning. Unlike usual object algorithms which use general...

Abstract Objective The aim of the study was to analyze self-monitoring ankle impedance measuring instrument for patients with heart failure and evaluate degree variation in normal young people three days. Methods We developed a portable based on AD5940 chip ADI. circuit composed programmable alternating current (AC) voltage generator, digital signal processor, microcontroller, related peripheral circuits. four-line body analysis measurement method used, which powered by two 1.5-V batteries,...

To address the issue of low segmentation accuracy for small objects in Mask Dino method, we propose an improved object model called FFMask Dino. Initially, introduce scaled cosine attention and log-cpb method into Swin Transformer backbone network. Subsequently, by adjusting network structure, enhance feature extraction process, which helps maintain generalization across different datasets reduces risk overfitting. Lastly, FFPN module to optimize pathways fusion transmission. The FPN...

Accurate prediction of multiaxial fatigue life was crucial for structural integrity assessment, yet the variability in material responses under complex loading paths made it challenging both classical and data-driven models to achieve high accuracy. To address this issue, a contrastive learning-based framework proposed study, enabling construction more generalized low-dimensional feature representations across different paths. This enhanced robustness without relying on mechanical...

3D skeleton-based action recognition and motion prediction are two essential problems of human activity understanding. In many previous works: 1) they studied tasks separately, neglecting internal correlations; 2) did not capture sufficient relations inside the body. To address these issues, we propose a symbiotic model to handle jointly; scales graphs explicitly among body-joints body-parts. Together, graph neural networks, which contain backbone, an action-recognition head,...

With the increasing availability of LCD displays and phone cameras in today's environment, screen-camera communication using dynamic barcode has emerged as a convenient infrastructure-free form to establish impromptu channel among mobile devices. Due short wavelengths narrow beams visible light, is highly directional, low-interference secure, which envisions wide range application scenarios. Conventional systems encode data bits with color barcodes, suffers from frame mixture problem caused...

Abstract Background Cone beam computed tomography (CBCT) image segmentation is crucial in prostate cancer radiotherapy, enabling precise delineation of the gland for accurate treatment planning and delivery. However, poor quality CBCT images poses challenges clinical practice, making annotation difficult due to factors such as noise, low contrast, organ deformation. Purpose The objective this study create a model label‐free target domain (CBCT), leveraging valuable insights derived from...

Reducing redundancy is crucial for improving the efficiency of video recognition models. An effective approach to select informative content from holistic video, yielding a popular family dynamic methods. However, existing methods focus on either temporal or spatial selection independently while neglecting reality that redundancies are usually and temporal, simultaneously. Moreover, their selected cropped with fixed shapes (<i>e.g</i>., temporally-cropped frames, spatially-cropped patches),...

Oracle Bone Inscriptions (OBI) are ancient hieroglyphs originated in China and considered one of the most famous writing systems world. Up to now, thousands OBIs have been discovered, which require deciphering by experts understand their contents. Experts typically need restore, classify, compare each character with previous inscriptions. Although existing research can assist these operations, performance falls short practical requirements. In this work, we propose OraclePoints framework,...

In this paper, we propose a novel general framework for tensor based null space affine invariants, namely, invariants (TNSI) with linear classifier high order data classification and retrieval. We first derive TNSI, which is perfectly invariant to multidimensional transformations due camera motions multiple motion trajectories in consecutive events. subsequently an efficient retrieval system relying on TNSI archiving searching events consisting of trajectories. The simulation results...

Wafer bin map (WBM) represents specific defect patterns that provide information for diagnosing root causes of low yield in semiconductor manufacturing. In practice, most engineers use subjective and time-consuming eyeball analysis to assess patterns. Given shrinking feature sizes, various types WBMs with different occur; therefore, relying on human vision judge become more complicated, inconsistent, unreliable. To bridge the gap, a system is proposed facilitating WBM extraction assisting...

Tremendous amounts of equipment parameters and sensor values were generated during modern semiconductor wafer manufacturing. These process data utilized for early diagnosis anomalies to prevent subsequent yield loss. However, from different steps machines highly customized that it is great difficulties traditional fault detection classification (FDC) analysis find a universal model identify excursions. In this paper, we present neural network method with deep convolutional variational...