Monocular dynamic reconstruction is a challenging and long-standing vision problem due to the highly ill-posed nature of the task. Existing approaches depend on templates, are effective only in quasi-static scenes, or fail to model 3D motion explicitly. We introduce a method for reconstructing generic dynamic scenes, featuring explicit, persistent 3D motion trajectories in the world coordinate frame, from casually captured monocular videos. We tackle the problem with two key insights: First, we exploit the low-dimensional structure of 3D motion by representing scene motion with a compact set of SE(3) motion bases. Each point's motion is expressed as a linear combination of these bases, facilitating soft decomposition of the scene into multiple rigidly-moving groups. Second, we take advantage of off-the-shelf data-driven priors such as monocular depth maps and long-range 2D tracks, and devise a method to effectively consolidate these noisy supervisory signals, resulting in a globally consistent representation of the dynamic scene. Experiments show that our method achieves state-of-the-art performance for both long-range 3D/2D motion estimation and novel view synthesis on dynamic scenes. Project Page: https://shape-of-motion.github.io/

A reference double-sampling phase-locked-loop-based (RDSPLL-based) multi-phase clock generator (MPCG) for DDR PHY is presented. The reference double-sampling architecture is utilized to achieve low phase noise. A CDAC-embedded voltage offset calibration is proposed to reduce jitter and reference spur, and a CMP-ADC hybrid phase detector is adopted to accelerate the locking process. Fabricated in 65nm, the proposed MPCG achieves better than 1° phase accuracy with a 100MHz reference clock. The reference spur is reduced from -56dBc to -80dBc and the locking time is reduced from 10.5us to 1.6us. The measured RMS jitter is 674fs at 2.4GHz with 3.43mW, yielding the FOMJitter-Nof -251.9dB.

This paper presents a reference double-sampling phase-locked loop (RDSPLL) that integrates frequency multiplication and octa-phase clock generation into a single system, significantly reducing power consumption. A differential ring oscillator (DRO) is employed to generate octa-phase clocks with high phase accuracy. The reference double-sampling technique extends the loop bandwidth, effectively suppressing phase noise from the ring oscillator and thereby reducing jitter. To achieve accurate and efficient phase error detection, we proposed a novel offset-compensated hybrid phase detector (OCH-PD), featuring an offset calibration and a comparator-ADC hybrid quantizer. The offset calibration utilizes the CDAC to dynamically compensate for the mismatch in double-sampling, improving jitter and spur performance. The hybrid quantizer supports dynamic mode switching based on different locking states: during the coarse frequency locking phase, it operates in the ADC mode to accelerate the locking process; once a stable lock is achieved, it switches to the comparator mode to enable low-power, high-speed quantization. Fabricated in a 65-nm CMOS process, the prototype achieves 674 fs RMS jitter at 2.4 GHz while consuming only 3.43 mW, resulting in a$\text {FOM}_{\text {Jitter-N}}$of -251.9 dB. With offset calibration, the reference spur at 100 MHz is suppressed from -56 dBc to -80 dBc, and the jitter is reduced from 1.42 ps to 674 fs. The locking time improves from$10.5~{\mu }$s to$1.6~{\mu }$s using the hybrid quantizer. The eight-phase accuracy remains better than 1° over the frequency range of 2-2.8 GHz.

We present Splat-Nav, a real-time robot navigation pipeline for Gaussian splatting (GSplat) scenes, a powerful new 3-D scene representation. Splat-Nav consists of two components: first, Splat-Plan, a safe planning module, and second, Splat-Loc, a robust vision-based pose estimation module. Splat-Plan builds a safe-by-construction polytope corridor through the map based on mathematically rigorous collision constraints and then constructs a Bézier curve trajectory through this corridor. Splat-Loc provides real-time recursive state estimates given only an RGB feed from an on-board camera, leveraging the point-cloud representation inherent in GSplat scenes. Working together, these modules give robots the ability to recursively replan smooth and safe trajectories to goal locations. Goals can be specified with position coordinates, or with language commands by using a semantic GSplat. We demonstrate improved safety compared to point cloud-based methods in extensive simulation experiments. In a total of 126 hardware flights, we demonstrate equivalent safety and speed compared to motion capture and visual odometry, but without a manual frame alignment required by those methods. We show online replanning at more than 2 Hz and pose estimation at about 25 Hz, an order of magnitude faster than neural radiance field-based navigation methods, thereby enabling real-time navigation.

Factored feature volumes offer a simple way to build more compact, efficient, and intepretable neural fields, but also introduce biases that are not necessarily beneficial for realworld data. In this work, we (1) characterize the undesirable biases that these architectures have for axis-aligned signals—they can lead to radiance field reconstruction differences of as high as 2 PSNR—and (2) explore how learning a set of canonicalizing transformations can improve representations by removing these biases. We prove in a simple two-dimensional model problem that a hybrid architecture that simultaneously learns these transformations together with scene appearance succeeds with drastically improved efficiency. We validate the resulting architectures, which we call TILTED, using 2D image, signed distance field, and radiance field reconstruction tasks, where we observe improvements across quality, robustness, compactness, and runtime. Results demonstrate that TILTED can enable capabilities comparable to baselines that are 2x larger, while highlighting weaknesses of standard procedures for evaluating neural field representations.

Microscopic hyperspectral image (MHSI) has received considerable attention in the medical field. The wealthy spectral information provides potentially powerful identification ability when combining with advanced convolutional neural network (CNN). However, for high-dimensional MHSI, the local connection of CNN makes it difficult to extract the long-range dependencies of spectral bands. Transformer overcomes this problem well because of its self-attention mechanism. Nevertheless, transformer is inferior to CNN in extracting spatial detailed features. Therefore, a classification framework integrating transformer and CNN in parallel, named as Fusion Transformer (FUST), is proposed for MHSI classification tasks. Specifically, the transformer branch is employed to extract the overall semantics and capture the long-range dependencies of spectral bands to highlight the key spectral information. The parallel CNN branch is designed to extract significant multiscale spatial features. Furthermore, the feature fusion module is developed to effectively fuse and process the features extracted by the two branches. Experimental results on three MHSI datasets demonstrate that the proposed FUST achieves superior performance when compared with state-of-the-art methods.

To a great extent, the quality of teaching determines the level of trained talents. Now it has entered the era of intelligent education, coupled with the rapid development of the Internet, has produced a large number of teaching data, which also makes the monitoring and evaluation of teaching quality become particularly difficult. In view of the above problems, this paper proposes the construction of big data monitoring platform for teaching quality under intelligent education. In this paper, the OPC UA unified architecture is used for communication between devices, and the information configuration is based on the spring boot framework to achieve data collection. Then, data processing is based on GRU neural network, and spark distributed computing framework is used to improve the efficiency of data operation. Finally, the monitoring effect is realized by constructing the evaluation system.

As one of the three provinces in Northeast China, Liaoning Province is one of the major industrial provinces. With the development of industrialization, the urban environmental pollution in Liaoning Province is becoming more and more serious, which has brought a huge impact on people's daily lives. In order to strengthen the monitoring of urban pollution, this paper proposes a study on the intelligent monitoring platform for urban pollution in Liaoning Province based on big data. This paper combines big data and intelligent monitoring technology to study the intelligent monitoring platform for urban pollution in Liaoning Province based on big data. The monitoring of air pollution in a city in Liaoning Province through simulation experiments verified the reliability of this platform. This paper finds that the average response time of the Liaoning Province's urban pollution intelligent monitoring platform based on big data proposed in this paper is 2.142 seconds, the transaction passing rate is 100%, and the platform performance test can pass. Through this platform, online real-time monitoring and remote control can be carried out in real time to grasp the dynamic changes of the environment, so that environmental problems can be found and treated in a timely manner, the environmental protection department's ability to prevent and control environmental pollution can be effectively improved, and it can also provide future urban pollution monitoring research.