Bird's Eye View (BEV) semantic segmentation is a critical task in autonomous driving. However, existing Transformer-based methods confront difficulties transforming Perspective (PV) to BEV due their unidirectional and posterior interaction mechanisms. To address this issue, we propose novel Bi-directional Early Interaction Transformers framework named BAEFormer, consisting of (i) an early-interaction PV-BEV pipeline (ii) bi-directional cross-attention mechanism. Moreover, find that the image...

Purpose: This study aims to investigate the effects of Huang Gan formula (HGF), a Chinese herbal prescription used for chronic kidney disease (CKD), on regulation gut microbiota and colonic microenvironment CKD. Methods: CKD rats were induced by 150 mg/kg adenine gavage 4 weeks, then orally treated with or without 3.6 g/kg 7.2 HGF 8 weeks. The renal function structure analyzed biochemical detection, hematoxylin eosin, Masson’s trichrome, Sirius red immunochemical staining. Average fecal...

Edge detection is a fundamental technique in various computer vision tasks. Edges are indeed effectively delineated by pixel discontinuity and can offer reliable structural information even textureless areas. State-of-the-art heavily relies on pixel-wise annotations, which labor-intensive subject to inconsistencies when acquired manually. In this work, we propose novel self-supervised approach for edge that employs multi-level, multi-homography transfer annotations from synthetic real-world...

In this paper, we propose a method for joint 2D segmentation and 2D-3D pose tracking. First, define novel energy functional which considers the discrimination between statistical appearance models coherence among neighboring pixels simultaneously. And then, particle filter-like stochastic optimization technique is adopted to solve functional, so that preferable initial value can be provided subsequent damped Newton method. Furthermore, an occlusion-aware updating strategy utilized models,...

Binary features have been incrementally popular in the past few years due to their low memory footprints and efficient computation of Hamming distance between binary descriptors. They shown with promising results on some real time applications, e.g., SLAM, where matching operations are relative few. However, computer vision, there many applications such as 3D reconstruction requiring lots local features. Therefore, a natural question is that feature still solution this kind applications? To...

High-definition (HD) map serves as the essential infrastructure of autonomous driving. In this work, we build up a systematic vectorized annotation framework (termed VMA) for efficiently generating HD large-scale driving scene. We design divide-and-conquer scheme to solve spatial extensibility problem generation, and abstract elements with variety geometric patterns unified point sequence representation, which can be extended most in VMA is highly efficient extensible, requiring negligible...

Deep learning has been successfully applied to single image super-resolution problem due its high data fitting ability. However, the trending of deeper layers and wider receptive field acquire better performance brings computation complexity serious information vanishing. To address this problem, we proposed a new Reconstructed DenseNets model for super-resolution. The basic idea behind is improve recent by modifying two core modules, dense blocks transition blocks, so that can emphasize...

A new method of three-dimensional (3D) image optical encryption based on phase-shifting compressive Fresnel holography (PSCFH) is presented. In the step, 3D object encrypted with random phase masks using interferometry to record three intensity images different phase-shifting. decryption three-step technique first used extract complex field at recording plane. Finally, reconstructed from recovered correct key. Numerical experiments demonstrate feasibility and robustness our proposed PSCFH method.

Image research has shown substantial attention in deblurring networks recent years. Yet, their practical usage real-world deblurring, especially motion blur, remains limited due to the lack of pixel-aligned training triplets (background, blurred image, and blur heat map) restricted information inherent images. This paper presents a simple yet efficient framework synthetic restore images using Inertial Measurement Unit (IMU) data. Notably, includes strategy for triplet generation,...

To explore the mass distribution characteristics of wet-mix shotcrete during spraying process and adhesion process, taking used for tunnel support as background, concrete spray in air attached to receiving surface were talked about by carrying out some tests at scale (spray scale). The differences between them compared relationship was analyzed. Besides, plastic viscosity regarded important factor, whose effect deeply investigated. results showed that three-dimensional shape impact section...

Introduction Motion blur, primarily caused by rapid camera movements, significantly challenges the robustness of feature point tracking in visual odometry (VO). Methods This paper introduces a robust and efficient approach for motion blur detection recovery blur-prone environments (e.g., with movements uneven terrains). Notably, Inertial Measurement Unit (IMU) is utilized detection, followed selection restoration strategy within frame sequence. It marks substantial improvement over...

Extending Bayesian optimization to batch evaluation can enable the designer make most use of parallel computing technology. Most current approaches artificial functions simulate sequential algorithm's behavior select a points for evaluation. However, as size grows, accumulated error introduced by these increases rapidly, which dramatically decreases efficiency algorithm. In this work, we propose simple and efficient approach extend Different from existing approaches, idea new is draw...

Yawning detection is actively used in multimedia applications such as driver fatigue assessment and status monitoring. However, the accuracy robustness of existing yawning detectors are limited due to variations environments (especially lights), facial expressions, confusion behaviours (e.g., talking eating). This paper introduces a transformer-based method, YawnNet, for accurate by leveraging spatial-temporal encoding local cues. In particular, YawnNet contains data processing stage with...

Estimating the 3D structure of drivable surface and surrounding environment is a crucial task for assisted autonomous driving. It commonly solved either by using sensors such as LiDAR or directly predicting depth points via deep learning. However, former expensive, latter lacks use geometry information scene. In this paper, instead following existing methodologies, we propose Road Planar Parallax Attention Network (RPANet), new neural network sensing from monocular image sequences based on...

Aiming at the great challenge for Computer Generated Hologram (CGH) duo to production of high spatial-bandwidth product (SBP) is required in real-time holographic video display systems. The paper based on point-cloud method and it takes advantage propagating reversibility Fresnel diffraction direction fringe pattern a point source, known as Gabor zone plate has spatial symmetry, so can be used basis fast calculation field CGH. A CGH novel look-up table (N-LUT) proposed, principle patterns...