Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks382 from public data
- An interpretable mortality prediction model for COVID-19 patients1,312
Overall, this Article suggests a simple and operable decision rule to quickly predict patients at the highest risk, allowing them to be prioritized and potentially reducing the mortality rate.
- Learning to Upsample by Learning to Sample767
Dysample, an ultra-lightweight and effective dynamic upsampler that requires no customized CUDA package and has much fewer parameters, FLOPs, GPU memory, and latency, outperforms other upsamplers across five dense prediction tasks.
- 300
- TasselNet: counting maize tassels in the wild via local counts regression network270
This is the first time that a plant-related counting problem is considered using computer vision technologies under unconstrained field-based environment and TasselNet can achieve robust in-field counting of maize tassels with a relatively high degree of accuracy.
- VisDrone-DET2020: The Vision Meets Drone Object Detection in Image Challenge Results263
Based on the selected 29 robust detection methods, the experimental results comprehensively are discussed, which shows the effectiveness of ensemble learning and data augmentation in drone captured object detection.
- 234
- A fast and robust local descriptor for 3D point cloud registration228
This paper proposes a novel local feature descriptor, called a local feature statistics histogram (LFSH), for efficient 3D point cloud registration, and an optimized sample consensus (OSAC) algorithm is developed to iteratively estimate the optimum transformation from point correspondences.
- 224
- P2B: Point-to-Box Network for 3D Object Tracking in Point Clouds191
This work first localize potential target centers in 3D search area embedded with target information, then point-driven 3D target proposal and verification are executed jointly, so the time-consuming 3D exhaustive search can be avoided.
- From Open Set to Closed Set: Counting Objects by Spatial Divide-and-Conquer191
S-DCNet achieves the state-of-the-art performance on three crowd counting datasets, a vehicle counting dataset (TRANCOS) and a plant counting datasets (MTC), and can generalize to open-set counts via S-DC.
- A2J: Anchor-to-Joint Regression Network for 3D Articulated Pose Estimation From a Single Depth Image190
The proposed 3D articulated pose estimation paradigm is different from the state-of-the-art encoder-decoder based FCN, 3D CNN and point-set based manners and 2D CNN is used as backbone network to drive A2J, without using time-consuming 3D convolutional or deconvolutional layers.
- TOLDI: An effective and robust approach for 3D local shape description171
Experimental results and comparisons with the state-of-the-arts validate the effectiveness, robustness, high efficiency, and overall superiority of the TOLDI method for local shape description.
- TasselNetv2: in-field counting of wheat spikes with context-augmented local regression networks159
TasselNetv2 for counting wheat spikes is described, which simultaneously addresses two important use cases in plant counting: improving the counting accuracy without increasing model capacity, and improving efficiency without sacrificing accuracy.
- Represent, Compare, and Learn: A Similarity-Aware Framework for Class-Agnostic Counting146
A similarity-aware CAC framework that jointly learns representation and similarity metric and extends BMNet to BMNet+ that models similarity from three aspects that significantly outperform state-of-the-art CAC approaches.
- NM-Net: Mining Reliable Neighbors for Robust Feature Correspondences140
This work proposes a hierarchical network named NM-Net with a series of graph convolutions that is insensitive to the order of correspondences that achieves the state-of-the-art performance on four datasets with various inlier ratios and varying numbers of feature consistencies.
- The All-Seeing Project: Towards Panoptic Visual Recognition and Understanding of the Open World133
The All-Seeing model (ASM), a unified framework for panoptic visual recognition and understanding, is developed with open-ended language prompts and locations, which allows it to generalize to various vision and language tasks with remarkable zero-shot performance.
- Automatic image-based detection technology for two critical growth stages of maize: Emergence and three-leaf stage133
A novel crops segmentation method (AP-HI) which is robust and not sensitive to the challenging variation of outdoor luminosity and complex environmental elements is proposed which can meet the demand for practical observation needed for agronomic modeling and in triggering action alerts to farmers.
- Action recognition for depth video using multi-view dynamic images130
This study extends dynamic imaging to the depth domain and proposes a novel CNN model to perform feature learning on multi-view dynamic images aimed at enhancing the tuning effectiveness on shallow convolutional layers by alleviating the gradient vanishing problem.
- 3DV: 3D Dynamic Voxel for Action Recognition in Depth Video127
3D dynamic voxel (3DV) is proposed as a novel 3D motion representation manner that is lightweight and effective for deep feature learning towards point set and split the depth video into temporal segments and encoded in 3DV integrally.
- 127
- MVSGaussian: Fast Generalizable Gaussian Splatting Reconstruction from Multi-View Stereo125
A new generalizable 3D Gaussian representation approach derived from Multi-View Stereo that can efficiently reconstruct unseen scenes with convincing generalizability, real-time rendering speed, and fast per-scene optimization.
- Crop segmentation from images by morphology modeling in the CIE L*a*b* color space117
A new morphology modeling method is utilized to establish the crop color model in the CIE L^*a^*b^* (or Lab for simplification) color space and to realize the crop image segmentation, demonstrating that this method is robust to the variation of illumination in the field and performed better than eight other approaches.
- 112
- Point-Query Quadtree for Crowd Counting, Localization, and More110
For the first time, a single counting model can address multiple crowd-related tasks across different learning paradigms and is implemented via data-dependent quadtree, where each querying point could split into four new points when necessary, thus enabling dynamic processing of sparse and dense regions.
- 110
- In-field automatic observation of wheat heading stage using computer vision109
Experimental results show that the proposed method significantly outperforms other existing methods with an average value of absolute error of 1.14 days on the test dataset, indicating that automatic observation is quite acceptable compared to manual observations.
- 102
- Deep Attention-Based Classification Network for Robust Depth Prediction98
This paper presents the deep attention-based classification (DABC) network for robust single image depth prediction, in the context of the Robust Vision Challenge 2018 (ROB 2018), and employs a soft-weighted sum inference strategy for the final prediction.
- SAPA: Similarity-Aware Point Affiliation for Feature Upsampling97
The key idea of the formulation is to generate similarity-aware kernels by comparing the similarity between each encoder feature point and the spatially associated local region of decoder features, which can function as a cue to inform the semantic cluster of upsampled feature points.
- TasselNetV2+: A Fast Implementation for High-Throughput Plant Counting From High-Resolution RGB Imagery92
By profiling the computational bottleneck, this work implements a fast version of a state-of-the-art plant counting model TasselNetV2 with several minor yet effective modifications, and encourages another promising paradigm termed object counting where plant counts are directly regressed from images, without detecting bounding boxes.
- 87
- Rice heading stage automatic observation by multi-classifier cascade based rice spike detection method86
Experiments demonstrated that the proposed method is an effective approach of automatic observation of the rice heading stage in paddy fields and can be utilized to replace the manual observation.
- Weighing Counts: Sequential Crowd Counting by Reinforcement Learning85
Inspired by scale weighing, this work proposes a novel 'counting scale' termed LibraNet where the count value is analogized by weight and shows that Libra net exactly implements scale weighing by visualizing the decision process howLibraNet chooses actions.
- In-field cotton detection via region-based semantic image segmentation85
A novel in-field cotton detection via region-based semantic image segmentation with two perspectives of observation, including unsupervised region generation and supervised semantic labeling prediction is reported on.
- When Unsupervised Domain Adaptation Meets Tensor Representations81
This work proposes a learning-based method to adapt the source and target tensor representations directly, without vectorization, and shows that this method outperforms the state-of-the-art on the task of cross-domain visual recognition in both efficacy and efficiency.
- 80
- 79
- Cloud Classification of Ground-Based Images Using Texture–Structure Features78
An algorithm is proposed that can capture both texture and structure information from a color sky image that combines a novel automatic block assignment method and a preprocessing color census transform.
- Rice Plant Counting, Locating, and Sizing Method Based on High-Throughput UAV RGB Images76
Results suggest that RiceNet can accurately and efficiently estimate the number of rice plants and replace the traditional manual method.
- 76
- Integration of the saliency-based seed extraction and random walks for image segmentation76
A novel automatic image segmentation method is proposed that combines the region saliency based on entropy rate superpixel (RSBERS) with the affinity propagation clustering algorithm to get seeds in an unsupervised manner, and uses random walks method to obtain the segmentation results.
- BokehMe: When Neural Rendering Meets Classical Rendering75
BokehMe is a hybrid bokeh rendering framework that marries a neural renderer with a classical physically motivated renderer that compares favorably against previous methods on both synthetic image data and real image data with predicted disparity.
- Maize tassels detection: a benchmark of the state of the art72
A comprehensive benchmark of state-of-the-art object detection and object counting methods, including Faster R-CNN, YOLOv3, FaceBoxes, RetinaNet, and the leading counting model of maize tassels—TasselNet is presented.
- Vegetation segmentation robust to illumination variations based on clustering and morphology modelling67
A new vegetation segmentation method based on Particle Swarm Optimisation (PSO) clustering and morphology modelling in CIE L∗a∗b∗ colour space is presented and it is demonstrated that the proposed method yielded the highest mean of segmentation qualities and lowest standard deviations of segmentsation qualities.
- 65
- 64
- 63
- Fine-grained maize tassel trait characterization with multi-view representations62
An automatic fine-grained machine vision system termed mTASSEL is developed that can serve the automatic growth stage detection, accurate yield estimation and machine detasseling, as well as the field-based phenotyping research.
- A real-time embedded architecture for SIFT62
This paper presents a low-cost embedded system based on a new architecture that successfully integrates FPGA and DSP that can detect SIFT feature and extract SIFT descriptor for detected features in real-time.
- A humanized anti-osteopontin antibody inhibits breast cancer growth and metastasis in vivo62
The data strongly support that OPN is a potential target for the antibody-based therapies of breast cancer and the humanized anti-OPN antibody hu1A12 may be a promising therapeutic agent for the treatment of human breast cancer.
- Rotational contour signatures for both real-valued and binary feature representations of 3D local shape61
A rotational contour signatures method for both real-valued and binary descriptions of 3D local shape and its seamless extension to binary representations to accelerate feature matching and reduce storage consumption is presented.
- Monocular Depth Estimation with Augmented Ordinal Depth Relationships59
It is shown that relative depth can be an informative cue for metric depth estimation and can be easily obtained from vast stereo videos, and a new “relative depth in stereo” (RDIS) dataset densely labeled with relative depths is introduced.
- Region-based colour modelling for joint crop and maize tassel segmentation59
Results show that the proposed method significantly outperforms other state-of-the-art approaches on tassel segmentation, with average precision of 74.3%, and achieves comparable performance of 77.8% on the traditional crop segmentation even with the naivest colour feature.
- 57
- Measuring Generalisation to Unseen Viewpoints, Articulations, Shapes and Objects for 3D Hand Pose Estimation Under Hand-Object Interaction57
A public challenge to evaluate the abilities of current 3D hand pose estimators (HPEs) to interpolate and extrapolate the poses of a training set, with dramatically improved accuracy over the baseline.
- When Epipolar Constraint Meets Non-local Operators in Multi-View Stereo53
A line-to-point non-local feature augmentation strategy that incorporates the Epipolar Transformer into a learning-based MVS baseline, named ET-MVSNet, which achieves state-of-the-art reconstruction performance on both the DTU and Tanks-and-Temples benchmark with high efficiency.
- 53
- 53
- AIM 2020 Challenge on Rendering Realistic Bokeh52
The proposed solutions significantly improved the baseline results, defining the state-of-the-art for practical bokeh effect rendering problem.
- A2J-Transformer: Anchor-to-Joint Transformer Network for 3D Interacting Hand Pose Estimation from a Single RGB Image49
The key idea is to equip A2J with strong local-global aware ability to well capture interacting hands' local fine details and global articulated clues among joints jointly to address 3D interacting hand pose estimation from a single RGB image.
- 49
- Multi-attribute statistics histograms for accurate and robust pairwise registration of range images49
A distinctive, robust and efficient 3D local shape descriptor called MaSH is proposed, and feature matching experiments show the over-all superiority of the proposed MaSH descriptor against the state-of-the-arts including the spin image, snapshots, THRIFT, FPFH, RoPS, LFSH and RCS descriptors.
- 49
- DoF-NeRF: Depth-of-Field Meets Neural Radiance Fields48
DoF-NeRF is introduced, a novel neural rendering approach that can deal with shallow DoF inputs and can simulate DoF effect and extends NeRF to simulate the aperture of lens following the principles of geometric optics.
- Crop feature extraction from images with probabilistic superpixel Markov random field48
A novel crop extraction method resistant to the strong illumination by using probabilistic superpixel Markov random field and capable of extracting the crop from the shadow regions, which can meet the requirement of practical uses in the authors' agriculture automatic vision system.
- Survey on depth and RGB image-based 3D hand shape and pose estimation47
This work states that the existence of complicated hand articulation, depth and scale ambiguities, occlusions, and the availability of large-scale annotated hand datasets and the rapid developments of deep neural networks (DNNs) means that accurate and rapid hand shape and pose estimation is needed.
- Entropic image thresholding based on GLGM histogram47
GLGM (gray-level & gradient-magnitude) histogram is proposed as a novel image histogram for thresholding and employs the Fibonacci quantized gradient magnitude to characterize spatial information effectively.
- Less is More45
This work proposes the frame masking network (FMNet), a spatial-temporal transformer network predicting the depth of masked frames based on their neighboring frames, which can learn intrinsic inter-frame correlations, which leads to consistency.
- Free4D: Tuning-Free 4D Scene Generation with Spatial-Temporal Consistency42
This work designs an adaptive guidance mechanism with a point-guided denoising strategy for spatial consistency and a novel latent replacement strategy for temporal coherence that enables real-time, controllable rendering in single-image-based 4D scene generation.
- High-Throughput Rice Density Estimation from Transplantation to Tillering Stages Using Deep Networks42
A deep learning-based approach called the Scale-Fusion Counting Classification Network (SFC2Net) that integrates several state-of-the-art computer vision ideas that addresses appearance and illumination changes by employing a multicolumn pretrained network and multilayer feature fusion to enhance feature representation.
- mCLOUD: A Multiview Visual Feature Extraction Mechanism for Ground-Based Cloud Image Categorization42
The experiments on a challenging cloud dataset termed the six-class Huazhong University of Science and Technology (HUST) cloud demonstrate that mCLOUD consistently outperforms other visual feature extraction algorithms in categorizing clouds based on ground-based images.
- 42
- Sparse-to-Dense Depth Completion Revisited: Sampling Strategy and Graph Construction41
This work proposes an end-to-end network with a graph convolution module that achieves not only state-of-the-art results for depth completion of indoor scenes but also better generalization ability than other competing methods.
- Towards Real-Time Eyeblink Detection in the Wild: Dataset, Theory and Practices40
A modified LSTM model able to capture the multi-scale temporal information is proposed, and a feature extraction approach that reveals the appearance and motion characteristics simultaneously is also proposed to verify eyeblink.
- Constraining Depth Map Geometry for Multi-View Stereo: A Dual-Depth Approach with Saddle-shaped Depth Cells39
An ideal depth geometry composed of Saddle-Shaped Cells is introduced, whose predicted depth map oscillates upward and downward around the ground-truth surface, rather than maintaining a continuous and smooth depth plane.
- 39
- Fast and Accurate Single-Image Depth Estimation on Mobile Devices, Mobile AI 2021 Challenge: Report39
This paper introduces the first Mobile AI challenge, where the target is to develop an end-to-end deep learning-based depth estimation solutions that can demonstrate a nearly realtime performance on smartphones and IoT platforms.
- 38
- Fine epitope mapping of humanized anti-IgE monoclonal antibody omalizumab37
The results indicated that omalizumab specifically bound to the synthesized peptide "(421)THPHLPRALMRS(432)" containing the (424)HLP(426) motif in IgE-Fc, thus providing insights into the structural basis for the mechanism of action of omalIZumab.
- 36
- 36
- 36
- DeepCotton: in-field cotton segmentation using deep fully convolutional network36
This work is the first to introduce deep learning to IFC segmentation, and greatly improves the property of the coarse result, with the average amplifications of 2.6%, 2.4% on accuracy and 8.5% on intersection over union for common scenarios and multiple plants.
- An improved MRF-based change detection approach for multitemporal remote sensing imagery35
An improved MRF based change detection approach for multitemporal remote sensing imagery is proposed that first finds edges in the difference image by using the line process, and the weights of MRF prior energy are adaptively adjusted by considering the gray level differences between neighboring pixels.
- 3D Cinemagraphy from a Single Image34
3D Cinemagraphy is presented, a new technique that mar-ries 2D image animation with 3D photography and proposes to bidirectionally displace the point cloud as per the scene flow and synthe-size novel views by separately projecting them into target image planes and blending the results.
- 33
- The effect of spatial information characterization on 3D local feature descriptors: A quantitative evaluation33
This study reveals that different spatial information encoding approaches would bring significant effect on a local shape descriptor with respect to its discriminative power, stability, compactness and efficiency.
- 32
- 32
- Type-2 fuzzy thresholding using GLSC histogram of human visual nonlinearity characteristics32
A new image thresholding method is presented using type-2 fuzzy sets based on GLSC histogram of human visual nonlinearity characteristics (HVNC) by embedding HVNC into GLSC Histogram.
- Fast Full-frame Video Stabilization with Iterative Optimization30
An iterative optimization-based learning approach using synthetic datasets for video stabilization, which consists of two interacting submodules: motion trajectory smoothing and full-frame outpainting, and formulate video stabilization as a problem of minimizing the amount of jerkiness in motion trajectories, which guarantees convergence with the help of fixed-point theory.
- 30
- 29
- 29
- New entropic thresholding approach using gray-level spatial correlation histogram29
The proposed entropic thresholding approach employing the gray-level spatial correlation (GLSC) histogram yields equivalent or even better results than 2-D ones while saving time remarkably and significantly outperforms Kapur's 1-D method without too much more time consumption generally.
- End-to-End Video Gaze Estimation via Capturing Head-Face-Eye Spatial-Temporal Interaction Context28
A new method, Multi-Clue Gaze (MCGaze), to facilitate video gaze estimation via capturing spatial-temporal interaction context among head, face, and eye in an end-to-end learning way, which has not been well concerned yet.
- 28
- 27
- Dynamic Color Transform Networks for Wheat Head Detection27
DCT plays an important role in the solution participating in the Global Wheat Challenge (GWC) 2021, where the solution ranks the first on the initial public leaderboard, and obtains the runner-up reward on the final private testing set, with an ADA of 0.695.
- 27
- 27
- Development and Validation of a Prognostic Risk Score System for COVID-19 Inpatients: A Multi-Center Retrospective Study in China27
A simple risk score has been validated to predict death in patients infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2); it has also been validated in independent cohorts.
- Performance Evaluation of 3D Correspondence Grouping Algorithms27
A thorough evaluation of several widely-used 3D correspondence grouping algorithms, motived by their significance in vision tasks relying on correct feature correspondences, and a summary of the merits/demerits of the evaluated algorithms from both performance and efficiency perspectives.
- 26
- 26
- A feasible approach for nonuniformity correction in IRFPA with nonlinear response26
Taking into account the nonlinearity of the photoresponse characteristic in infrared focal plane array, a feasible approach for nonuniformity correction of IRFPA's is proposed and the correction formula deduced is the best approximation polynomial of the analytic formula through theoretical analysis.
- DyBluRF: Dynamic Neural Radiance Fields from Blurry Monocular Video25
DyBluRF is proposed, a dynamic radiance field approach that synthesizes sharp novel views from a monocular video affected by motion blur that outperforms existing approaches in generating sharp novel views from motion-blurred inputs while maintaining spatial-temporal consistency of the scene.
- 25
- Tempo-Spatial Variation of Vegetation Coverage and Influencing Factors of Large-Scale Mining Areas in Eastern Inner Mongolia, China25
The vegetation coverage towards increasing in eastern Inner Mongolia decreased in the order of Tongliao > Hinggan League > Chifeng > Hulunbuir > Xilingol over the past 34 years, and vegetation change in Shengli No. 1 had a significant correlation with precipitation and human activities after mining.
- 24
- Make-It-4D: Synthesizing a Consistent Long-Term Dynamic Scene Video from a Single Image24
This work presents Make-It-4D, a novel method that can generate a consistent long-term dynamic video from a single image and can be training-free which saves a significant amount of training time.
- 24
- Learning to fuse local geometric features for 3D rigid data matching24
This paper shows that more compact and distinctive representations can be achieved by optimizing a neural network model under the triplet framework that non-linearly fuses local geometric features in Euclidean spaces by trained by an improved triplet loss function that fully leverages all pairwise relationships within the doublet.
- Interactive color image segmentation via iterative evidential labeling24
The proposed interactive color image segmentation method makes use of the conception of Markov random fields and D–S evidence theory to obtain segmentation results by considering both likelihood information and priori information under Bayesian framework.
- An image-based approach for automatic detecting tasseling stage of maize using spatio-temporal saliency24
A new method called Spatio-temporal Saliency Mapping is proposed to highlight the ear while suppress the background, which significantly improve the detection performance of the tasseling stage of maize.
- 23
- 23
- Towards Robust Monocular Depth Estimation: A New Baseline and Benchmark22
This work delves deeper into data augmentation and proposes utilizing strong data augmentation techniques for robust depth estimation, and introduces 3D-aware defocus blur in addition to seven 2D data augmentations to demonstrate better generalization and robustness than previous methods.
- From Open Set to Closed Set: Supervised Spatial Divide-and-Conquer for Object Counting22
The idea of spatial divide-and-conquer (S-DC) that transforms open-set counting into a closed set problem is introduced and implemented by a novel Supervised Spatial Divide-and-Conquer Network (SS-DCNet).
- 22
- A Compound Approach for Monthly Runoff Forecasting Based on Multiscale Analysis and Deep Network with Sequential Structure22
A novel compound approach combining time-varying filtering-based empirical mode decomposition (TVFEMD), sample entropy (SE)-based subseries recombination, and the newly developed deep sequential structure incorporating convolutional neural network (CNN) into a gated recurrent unit network (GRU) is proposed for monthly runoff forecasting.
- 22
- 21
- 3D Multi-frame Fusion for Video Stabilization21
A novel framework for video stabilization that integrates 3D multi-frame fusion through volume rendering, and introduces the Adaptive Ray Range (ARR) module to integrate depth priors, adaptively defining the sampling range for the projection process.
- ECML: An Ensemble Cascade Metric-Learning Mechanism Toward Face Verification21
The experimental results demonstrate that EC-RMML is superior to state-of-the-art metric-learning methods for face verification, and the proposed ECML mechanism is also applicable to other metric- learning approaches.
- Towards fine-grained maize tassel flowering status recognition: Dataset, theory and practice21
An automatic maize tassel flowering status recognition method via the computer vision technology is proposed that goes beyond the state-of-the-art by large margins and is made available online.
- Local phase quantization plus: A principled method for embedding local phase quantization into Fisher vector for blurred image recognition21
Local phase quantization plus ( LPQ + ), which embeds LPQ into Fisher vector (FV) to acquire mid-level blurred image representation under the bag-of-words (BoW) model, significantly outperforms LPQ and other well-established visual features in discriminative power.
- 20
- Expression, purification, and characterization of an immunotoxin containing a humanized anti-CD25 single-chain fragment variable antibody fused to a modified truncated Pseudomonas exotoxin A20
A recombinant immunotoxin CD25-PE38KDELKQK containing humanized anti-CD25 single-chain antibody genetically fused to PE38kDELkQK is constructed and expressed in Escherichia coli and shown to be much weaker in inducing VLS in mice than wt.
- 20
- 19
- Defocus to focus: Photo-realistic bokeh rendering by fusing defocus and radiance priors19
This work presents a Defocus to Focus (D2F) framework to learn realistic bokeh rendering by fusing defocus priors with the all-in-focus image and by implementing radiance priors in layered fusion, and introduces defocus hallucination to integrate depth by learning to focus.
- Performance Evaluation of Crop Segmentation Algorithms19
A new classification strategy of object segmentation is proposed by dividing the algorithms into pixel-based and region-based approaches at first, and then systematically evaluating various crop segmentation methods with a unified data benchmark and four common metrics.
- FADE: A Task-Agnostic Upsampling Operator for Encoder–Decoder Architectures18
This paper presents FADE, a novel, plug-and-play, lightweight, and task-agnostic upsampling operator by fusing the assets of decoder and encoder features at three levels, and demonstrates robust feature upsampling on both region- and detail-sensitive tasks successfully.
- Unsupervised domain adaptation for in-field cotton boll status identification18
A novel NCA-based unsupervised domain adaptation method termed NCADA, which includes three procedures: feature extraction using a deep CNN, feature transformation matrix generation, and target label inference, is proposed, demonstrating that NCADA can be a useful tool to replace manual observation and standard classification methods.
- A New Dew and Frost Detection Sensor Based on Computer Vision18
A novel method for indirectly detecting dew and frost based on computer vision is proposed, which is simple, inexpensive, and only requires images of several glass substrates near the underlying surface.
- TacoDepth: Towards Efficient Radar-Camera Depth Estimation with One-stage Fusion17
TacoDepth is proposed, an efficient and accurate Radar-Camera depth estimation model with one-stage fusion designed to capture and integrate the graph structures of Radar point clouds, delivering superior model efficiency and robustness without relying on the intermediate depth results.
- 17
- 17
- Change detection of multispectral remote-sensing images using stationary wavelet transforms and integrated active contours17
Experimental results obtained on multispectral remote-sensing images confirm the effectiveness of the proposed change-detection approach, although the time consumption of the approach is somewhat high.
- 16
- 16
- Two-dimensional subspace alignment for convolutional activations adaptation16
A two-dimensional subspace alignment (2DSA) approach based on 2D principal component analysis (PCA) to better adapt convolutional activations and introduces within- and between-class domain divergence measures to characterize the class-level differences.
- 15
- 14
- Diffusion-Augmented Depth Prediction with Sparse Annotations14
A supervised framework termed Diffusion-Augmented Depth Prediction (DADP), which leverages the structural characteristics of diffusion model to enforce depth structures of depth models in a plug-and-play manner and achieves significant improvements in depth structures and robustness.
- Real-time Multi-person Eyeblink Detection in the Wild for Untrimmed Video14
A large-scale dataset termed MPEblink that involves 686 untrimmed videos with 8748 eyeblink events is proposed under multi- person conditions and a real-time multi-person eyeblink detection method is proposed that outperforms existing approaches by large margins and with a high inference speed.
- Ranking-Based Salient Object Detection and Depth Prediction for Shallow Depth-of-Field14
A label-guided ranking loss for both salient object detection and depth prediction is introduced and a SOD and depth-aware blur rendering method is introduced to generate shallow DoF images.
- 14
- 14
- 14
- 14
- Learning Second-Order Attentive Context for Efficient Correspondence Pruning13
This paper first extends the attentive context to the first-order attentive context and then introduces the idea of attention in attention (ANA) to model second- order attentive context for correspondence pruning, and derives two approximate formulations for the naive implementation of second-order attention to optimize the cubic complexity to linear complexity.
- 13
- LRF-Net: Learning Local Reference Frames for 3D Local Shape Description and Matching13
This paper presents the first attempt to learn an LRF via a Siamese network that needs weak supervision only and argues that each neighboring point in the local surface gives a unique contribution to LRF construction and measures such contributions via learned weights.
- 13
- 13
- Forward-looking infrared target recognition based on histograms of oriented gradients13
A target recognition algorithm based on histograms of oriented gradients (HOG) which evaluates normalized local histogram of image gradient orientations in a dense grid and adopts linear SVM trained for a binary object/non-object classifier to detect the object in the forward-looking infrared (FLIR) images.
- 13
- Light-X: Generative 4D Video Rendering with Camera and Illumination Control12
This work presents Light-X, a video generation framework that enables controllable rendering from monocular videos with both viewpoint and illumination control, and introduces Light-Syn, a degradation-based pipeline with inverse-mapping that synthesizes training pairs from in-the-wild monocular footage.
- Exploring Contextual Attribute Density in Referring Expression Counting12
This work proposes a U- shape CAD estimator in which referring expression and multi-scale visual features from GroundingDINO can interact with each other and can effectively encode CAD, which is subsequently decoded via a novel attention procedure with CAD-refined queries.
- 12
- 12
- Image segmentation using modified SLIC and Nyström based spectral clustering12
An efficient two-stage image segmentation method is proposed which takes advantage of modified SLIC segmentation and Nystrom based spectral clustering and the memory requirement and the computational complexity are significantly reduced.
- 12
- 11
- Exposure Completing for Temporally Consistent Neural High Dynamic Range Video Rendering11
This paper proposes a novel paradigm to render HDR frames via completing the absent exposure information, hence the exposure information is complete and consistent, and benefits the fusing process for HDR results, reducing noise and ghosting artifacts therefore improving temporal consistency.
- 11
- 11
- 11
- 11
- 11
- Geometry-aware Reconstruction and Fusion-refined Rendering for Generalizable Neural Radiance Fields10
This work introduces an Adaptive Cost Aggregation (ACA) approach to amplify the contribution of consistent pixel pairs and suppress inconsistent ones, and introduces a Spatial-View Ag-gregator (SVA) to incorporate 3D context into descriptors through spatial and inter-view interaction.
- NVDS+: Towards Efficient and Versatile Neural Stabilizer for Video Depth Estimation10
NVDS is introduced that stabilizes inconsistent depth estimated by various single-image models in a plug-and-play manner and a bidirectional inference strategy is designed to improve consistency by adaptively fusing forward and backward predictions.
- 10
- A Self Training Mechanism With Scanty and Incompletely Annotated Samples for Learning‐Based Cloud Detection in Whole Sky Images10
The proposed self‐training mechanism can outperform unsupervised methods and achieve comparable results with fully supervised learning methods but using much fewer annotated samples.
- 10
- 10
- 9
- 9
- 9
- 9
- SimHMR: A Simple Query-based Framework for Parameterized Human Mesh Reconstruction9
This work proposes a simple query-based framework, dubbed SimHMR, for parameterized human mesh reconstruction, which effectively removes the need for hand-crafted intermediate representation and reasoning pipeline and significantly improves the efficiency and robustness in extreme scenarios, e.g., occlusions.
- 9
- 9
- 9
- 9
- Li Yan et al. reply9
To understand the differences in the performance of the model among these datasets and those in ref. 1, the distributions of the three key biomarkers were compared and showed that the data distributions for all three biomarkers, for Tongji and the other hospitals, are statistically different.
- 9
- Recognition of blurred faces using local phase pattern9
The results show that LPP achieves a promising recognition rate, even when the amount of blur significantly increases, and is assessed in comparison with LBP and local phase quantisation.
- Identity-Preserving Image-to-Video Generation via Reward-Guided Optimization8
Identity-Preserving Reward-guided Optimization (IPRO) is proposed, a novel video diffusion framework based on reinforcement learning to enhance identity preservation and a novel facial scoring mechanism that treats faces in ground-truth videos as facial feature pools, providing multi-angle facial information to enhance generalization.
- SRefiner: Soft-Braid Attention for Multi-Agent Trajectory Refinement8
Inspired by braid theory, a novel trajectory refinement approach, Soft-Braid Refiner (SRefiner), guided by the soft-braid topological structure of trajectories using Soft-Braid Attention is proposed, establishing a new state-of-the-art in trajectory refinement.
- Self-Supervised Class-Agnostic Motion Prediction with Spatial and Temporal Consistency Regularizations8
The feasibility of self-supervised motion prediction with only unlabeled Li-DAR point clouds is explored and the significant superiority of this approach over the state-of-the-art self-supervised methods is demonstrated.
- First Arrival Picking on Microseismic Signals Based on K-Means with a ReliefF Algorithm8
This study proposed a first arrival picking method for application to microseismic data with a low signal-to-noise ratio (SNR), which consisted of two steps: feature selection and clustering.
- 8
- 8
- Enhancement of X-ray angiogram images by adaptive anisotropic diffusion8
A new angiogram image enhancement algorithm based on anisotropic diffusion that can adaptively choose the conductance parameter, which has great significance on the whole diffusion, is proposed.
- DoF-Gaussian: Controllable Depth-of-Field for 3D Gaussian Splatting7
DoF-Gaussian is introduced, a controllable depth-of-field method for 3D-GS that combines a lens-based imaging model based on geometric optics principles to control DoF effects and applies defocus-to-focus adaptation to minimize the gap in the circle of confusion.
- S-DyRF: Reference-Based Stylized Radiance Fields for Dynamic Scenes7
S-DyRF is presented, a reference-based spatio-temporal stylization method for dynamic neu-ral radiance fields that generates temporal pseudo-references from the given stylized reference to facilitate the propagation of style infor-mation from the reference to the entire dynamic 3D scene.
- 7
- Learning Probabilistic Coordinate Fields for Robust Correspondences7
The interpretable confidence map predicted by PCF-Net can be leveraged to other novel applications from texture transfer to multi-homography classification, and can be used as a plug-in to existing correspondence-dependent approaches.
- 7
- 7
- Point-and-Shoot All-in-Focus Photo Synthesis From Smartphone Camera Pair7
Benefiting from the multi-camera module in modern smartphones, a new task of AIF synthesis from main (wide) and ultra-wide cameras is introduced, with results showing that the solution, termed EasyAIF, can produce high-quality AIF photos and outperforms strong baselines quantitatively and qualitatively.
- 3D Correspondence Grouping with Compatibility Features7
Comparisons with nine state-of-the-art methods on four benchmarks demonstrate that: 1) CF is distinctive, robust, and rotation-invariant; 2) the CF-based method achieves the best overall performance and holds good generalization ability.
- Multispectral Semantic Land Cover Segmentation From Aerial Imagery With Deep Encoder–Decoder Network7
- Hand pose estimation in depth image using CNN and random forest7
A two-stage framework which joint with CNN and Random Forest to boost the performance of hand pose estimation and proposes a pyramid feature which merges the information flow of the CNN.
- 7
- Real-time detecting system for infrared small target7
An infrared small target detecting parallel system is constituted, which makes better use of the advantages of DSP, FPGA and ASIC while avoiding their disadvantages.
- Instance Consistency Regularization for Semi-Supervised 3D Instance Segmentation6
A novel self-training network InsTeacher3D is proposed to explore and exploit pure instance knowledge from unlabeled data and significantly outperforms prior state-of-the-art semi-supervised approaches on large-scale datasets.
- In-Context Matting6
In-context matting is introduced, a novel task setting of image matting that enables automatic alpha estimation on a batch of target images of the same foreground category, without additional auxiliary input, and can make full use of reference context to generate accurate target alpha mattes.
- 6
- A2B: Anchor to Barycentric Coordinate for Robust Correspondence6
This study demonstrates that geometric-invariant coordinate representations, such as barycentric coordinates, can significantly reduce mismatches between features and introduces Degree, a novel anchor-to-barycentric (A2B) coordinate encoding approach, which generates multiple affine-invariant correspondence coordinates from paired images.
- Analysis of load and adaptability of disc cutters during shield tunneling in soft–hard varied strata6
- 6
- 6
- Supervised guiding long-short term memory for image caption generation based on object classes6
A supervised guiding Long Short Term Memory model based on object classes, named S-gLSTM for short, that uses the object detection results from R-FCN as supervisory information with high confidence, and updates the guidance word set by judging whether the last output matches thesupervisory information.
- 3D reconstruction from non-uniform point clouds via local hierarchical clustering6
Experimental results on real-world scanned point clouds validate the effectiveness of the proposed local hierarchical clustering (LHC) method from both qualitative and quantitative aspects.
- 6
- 6
- Robust key point descriptor for multi-spectral image matching6
Experiments demonstrate that the proposed key point descriptor for multi-spectral image matching can yield significantly better results than some state-of- the-art descriptors.
- A variational change detection method for multitemporal SAR images6
A variational model for change detection in multitemporal synthetic aperture radar (SAR) images that is a generalization of ACM under the assumption that the image data fits the Generalized Gaussian Mixture (GGM) model.
- An automatic detection method to the field wheat based on image processing6
A novel two-stage wheat image segmentation method that adjusts several key thresholds which will be used in segmentation stage to achieve the best segmentation results, and counts these thresholds.
- Automatic measurement of crops canopy height based on monocular vision6
A novel method is proposed, that is, to measure the canopy height indirectly by the solving algorithm for the actual height of the vertical objects (SAAH) with the help of the intelligent sensor device.
- An incremental learning algorithm based on Support Vector Machine for pattern recognition6
A SVM's incremental learning algorithm based on the filtering fixed partition of the data set and generalizes it to the "Multiclass problem" algorithm by the One-vs-One method to greatly improve the efficiency of SVM learning.
- Moving object detection and tracking in video surveillance system6
This paper presents solutions to some problems in real-time video surveillance system such as robust object detection in illumination mutation situation and merge/split event in multi-target tracking by combining Gaussian Mixture Model algorithm with temporal frame difference.
- A New Piecewise Approach for Nonuniformity Correction in IRFPA6
A novel piecewise linear connection scheme for nonuniformity correction in IRFPA which tries to find the n-segment piecewise straight-line which strives for going near (not always through) all available calibration points, rather than just passes through some n + 1 calibration points.
- 5
- iControl3D: An Interactive System for Controllable 3D Scene Generation5
iControl3D is presented, a novel interactive system that empowers users to generate and render customizable 3D scenes with precise control and enables users to build a radiance field of their scene online and navigate the entire scene.
- Video Bokeh Rendering: Make Casual Videography Cinematic5
The Video Bokeh Renderer (VBR), the model designed specifically for video bokeh rendering, is proposed and the first Synthetic Video Bokeh (SVB) dataset is introduced, synthesizing authentic bokeh effects using ray-tracing techniques.
- Design What You Desire: Icon Generation from Orthogonal Application and Theme Labels5
This paper proposes IconGAN composed of a conditional generator and dual discriminators with orthogonal augmentations, and a contrastive feature disentanglement strategy is further designed to regularize the feature space of the two discriminators.
- Background-Aware Domain Adaptation for Plant Counting5
It is shown that BADA can easily fit into object counting models to improve the cross-domain plant counting performance, especially on background areas, and can work with adversarial training strategies to further enhance the robustness of counting models against the domain gap.
- 5
- 5
- 5
- Action recognition in depth video from RGB perspective: A knowledge transfer manner5
This paper proposes a method for human action recognition from RGB video to Depth video using domain adaptation, where learned feature from RGB videos is used to do action recognition for depth videos.
- 5
- 5
- Digital implementation of nonuniformity correction for IRFPAs5
Measured results show that digital processing system designed by us may achieve demand of real-time nonuniformity correction based the scene for small IRFPAs.
- Explore More, Learn Better: Parallel MLLM Embeddings under Mutual Information Minimization4
This paper tailor one Parallel Decoupling Framework (PDF) for multimodal embedding learning, by utilizing the proprietary steerability of MLLMs, i.e., their ability to flexibly generate quite differentiated response under explicit instructions.
- WildAvatar: Learning In-the-wild 3D Avatars from the Web4
This paper curates WildAvatar, a web-scale in-the-wild human avatar creation dataset extracted from YouTube, with 10,000+ different human subjects and scenes, and demonstrates the quality and generalizability of avatar creation methods on WildAvatar.
- Dual-Camera All-in-Focus Neural Radiance Fields4
The first framework capable of synthesizing the all-in-focus neural radiance field (NeRF) from inputs without manual refocusing is presented, which can produce high-quality all-in-focus novel views and compares favorably against strong baselines quantitatively and qualitatively.
- 4
- 4
- 4
- 4
- 4
- 4
- Abrupt-motion-aware lightweight visual tracking for unmanned aerial vehicles4
The proposed abrupt-motion-aware visual tracking method is a lightweight tracker for small UAVs without the deep learning framework and an adaptive model update strategy is designed to alleviate the model noise caused by the short-term occlusion.
- LiDAR-Stereo4
Novel convolutional neural network architecture with two novel units: monocular completion unit and stereo fusion unit is designed, and the complementary properties of sparse LiDAR and dense stereo images are fully explored in a boosting manner.
- 4
- 4
- 4
- Specularity-invariant crop extraction with probabilistic super-pixel Markov random field4
This paper proposes a specularity-invariant crop extraction method using probabilistic super-pixel markov random field (MRF), which outperforms the other commonly used extraction methods in yielding highest performance with the lowest standard deviation.
- 4
- A study on the difficulty prediction for infrared target recognition4
A novel difficult prediction scheme for infrared building target recognition that can predict the difficulty of recognizing a designated target in advance in advance, which is desirable in infrared building recognition.
- A method of frost observation based on intensity changing regularity simulation and texture analysis4
The method can get high detection accuracy in the different kinds of continuous changing environment and the characteristic of frost surface texture is also well described by texture analysis based on texture descriptor.
- 4
- Implementation of large kernel 2-D convolution in limited FPGA resource4
An FIFO type line delayer is presented to serve as the data buffer for convolution to reduce the data fetching operation and a finite state machine is applied to control the reuse of multipliers and adders arrays.
- 3
- Generative Photographic Control for Scene-Consistent Video Cinematic Editing3
This paper proposes CineCtrl, the first video cinematic editing framework that provides fine control over professional camera parameters (e.g., bokeh, shutter speed), and introduces a decoupled cross-attention mechanism to disentangle camera motion from photographic inputs, allowing fine-grained, independent control without compromising scene consistency.
- Dynamic View Synthesis From Small Camera Motion Videos3
This work proposes a novel Distribution-based Depth Regularization (DDR) that ensures the rendering weight distribution to align with the true distribution and incorporates camera parameter learning during training to enhance the robustness of the model to camera parameters.
- 3
- 3
- 3
- 3
- 3
- Video stabilization based on image registration3
The quantitative and qualitative evaluations show that the proposed registration-based video stabilization method can produce comparable results with the state-of-the-art methods and achieve more stable performance on challenging situations.
- Machine learning discovery of distinguishing laboratory features for severity classification of COVID‐19 patients3
The discovered top 30 laboratory features can be used for greater differentiation to serve as an essential supplement to current guidelines, thus creating a more comprehensive assessment of COVID‐19 cases during the early stages of infection.
- 3
- 3
- Pose estimation for non-cooperative targets using 3D feature correspondences grouped via local and global constraints3
This work employs a combination of local and global constraints to remove the outliers in initial correspondences and proves that the quality of the correspondence set has been greatly improved by the method and the pose can be accurately estimated.
- Depth image super-resolution via semi self-taught learning framework3
This paper presents a new semi self-taught learning application framework for enhancing resolution of depth maps without making use of ancillary color images data at the target resolution, or multiple aligned depth maps.
- 3
- A FPGA-based architecture for real-time image matching3
This paper proposes a single FPGA-based image matching system, which consists of SIFT feature detection, BRIEF descriptor extraction and BriEF matching, which optimizes theFPGA architecture for the SIFTfeature detection to reduce the FFPA resources utilization.
- 3
- Forward-looking infrared 3D target tracking via combination of particle filter and SIFT3
A new method of estimating the affine transformation matrix parameters based on Monte Carlo methods of particle filter for robust tracking of 3D target in forward-looking infrared (FLIR) image is proposed.
- Classification method for aerial LiDAR data based on Markov random field3
Experimental results show the proposed method can classify the LiDAR data properly and is implemented by maximising the marginal probabilities, inferred from a Markov random field which models the neighbourhood and altitude difference between surface patches extracted from LiDar points.
- 3D target tracking in infrared imagery by SIFT-based distance histograms3
This paper proposes a novel algorithm for 3D target tracking in FLIR image sequences that uses SIFT keypoints detected in consecutive frames for point correspondence and efficiently improves the tracking performance and accuracy.
- 3
- 3
- Knowledge-based recognition algorithm for long-range infrared bridge images3
A new knowledge-based recognition algorithm that first detects the candidate bridge sub-regions and then focuses on them, so the false objects are excluded and eventually real target is found.
- 2
- 2
- 2
- Deblur-Avatar: Animatable Avatars from Motion-Blurred Monocular Videos2
This work introduces a novel framework for modeling high-fidelity, animatable 3D human avatars from motion-blurred monocular video inputs that significantly outperforms existing methods in rendering quality and quantitative metrics, producing sharp avatar reconstructions and enabling real-time rendering under challenging motion blur conditions.
- 2
- 2
- Box-DETR: Understanding and Boxing Conditional Spatial Queries2
This work proposes Box Agent to condense the box into head-specific agent points, and argues that the explicit use of the entire box information in cross-attention matters, which leads to not only faster convergence but also improved detection performance.
- On Point Affiliation in Feature Upsampling2
It is shown that an upsampled point can resort to its low-res decoder neighbors and high-res encoder point to reason the affiliation, conditioned on the mutual similarity between them, and it is proved that such kernels encourage not only semantic smoothness but also boundary sharpness.
- 2
- 2
- 2
- 2
- LPQ++: A discriminative blur-insensitive textural descriptor with spatial-channel interaction2
The main idea is to establish spatial-channel interactions between the normalized blur-insensitive feature maps yielded by a short-term Fourier transform (STFT) to enhance the descriptive power while maintaining the insensitivity to blur.
- 2
- Online Update Siamese Network for Unmanned Surface Vehicle Tracking2
An online update siamese network is proposed to address the temporal variation of target appearance in visual tracking for unmanned surface vehicle (USVs) and can effectively update the tracking template from previous frames.
- Real-Time Stereo Matching System2
This paper improves the traditional stereo matching method SGM so that it is able to meet the real-time requirements and gets great balance between the speed and the accuracy.
- 2
- Structured output tracking guided by keypoint matching2
To overcome the tracked point drifting, structured output tracking is used to track the target point with its surrounding information based on Haar-like features and is able to outperform state-of-the-art keypoint-based trackers.
- Refine BING using effective cascade ranking2
An effective and efficient cascade ranking method is proposed to refine BING and the concept of scale-sets histogram is novelly introduced, which helps to analyze the potential sizes of the objects.
- Effective 3D infrared image target tracking via framework of particle filter and FAST2
An effective tracking framework based on FAST BRIEF and Particle Filter (FBPF) is proposed which provides a convenient framework for estimating and propagating the posterior probability density function of state variables regardless of the underlying distribution through a sequence of prediction and update steps.
- [Investigation on schistosomiasis cognitive levels of people from Chaohu area].2
It is necessary for people from Chaohu to carry out various forms of health education of schistosomiasis prevention and control before the water transfer project operation.
- Blurred face recognition by fusing blur-invariant texture and structure features2
The texture and structure blur-invariant features are extracted and fused to generate a more complete description on blurred image and the improvements and proposition can have a good performance in blurred face recognition.
- Object matching task-oriented image quality assessment2
A system of image quality assessment related to object matching task is constructed and Gray level correlation matching and histograms of oriented gradients -matching are demonstrated and their experimental result shows that the image quality Assessment system works effectively.
- Color image segmentation using watershed and Nyström method based spectral clustering2
A novel two-stage color image segmentation method that uses vector gradient approach to detect color image gradient information, and watershed transformation to get the pre-segmentation result and Nyström extension based spectral clustering is used toget the final result.
- 2
- 2
- Automatic 3-D building reconstruction method based on singular point detection of 2-D derived function2
A new automatic building reconstruction method is proposed, which can solve the complex building reconstruction problem effectively and conveniently, and has excellent universality, besides a small calculating amount compared with traditional methods.
- 2
- Synthetic-to-Real Translation for Class-Agnostic Motion Prediction1
This work proposes a novel approach integrating a motion knowledge translation framework with two key components: 1) objectness-aware motion prediction, which explicitly models the joint distribution of motion patterns and objectness priors to improve domain-invariant feature learning, and 2) objectness-aided motion enhancement, a motion label refinement mechanism that leverages learned objectness priors to filter motion noise.
- Dynamic Beauty is Easy to Find: A Large-Scale Composition-Aware Dataset and an End-to-End Framework for Video Reframing1
Experiments show the proposed fully automated composition-aware dataset using vision-language models and image composition assessment models outperforms others in efficiency and effectiveness, proving that composition awareness and end-to-end modeling are critical for video reframing.
- MuGS: Multi-Baseline Generalizable Gaussian Splatting Reconstruction1
This work presents Multi-Baseline Gaussian Splatting (MuGS), a generalized feed-forward approach for novel view synthesis that effectively handles diverse baseline settings, including sparse input views with both small and large baselines, and proposes a projection-and-sampling mechanism for deep depth fusion.
- 1
- 1
- 1
- Real-time Embedded Demo System for Fall Detection under 15W Power1
An embedded demo system for real-time spatial-temporal fall detection on demo video is proposed, built upon MicroSoft Kinect V2 for 3D visual sensing, which is insensitive to illumination change, with end-to-end running capacity.
- 1
- 1
- 1
- Limited Receptive Field Network for Real-Time Driving Scene Semantic Segmentation1
Inspired by the mechanism of human eyes, a novel Limited Receptive Field Network (LRFNet) is proposed which achieves a good balance between the segmentation speed and accuracy and is superior to many state-of-the-art methods on two standard benchmarks.
- Video Object Detection and Segmentation Based on Proposal Boxes1
A new method to detect and segment foreground object in video automatically automatically and report comparable results with state-of-the-art works on challenging benchmark dataset is proposed.
- 1
- 1
- ATR algorithm performance evaluation based on the simulation image and real image1
This paper proposes a method to establish the relation model between image quality characteristic and ATR algorithm based on SVM classifier and uses the model to evaluate algorithm’s performance.
- 1
- 1
- 1
- 1
- Image fusion algorithm using nonsubsampled contourlet transform1
Experiment demonstrates that NSCT could preserve edge significantly and the fusion rule based on region segmentation performances well in local-contrast enhancement.
- 3-D Reconstruction of Blood Vessels Skeleton Based on Neural Network1
An efficient 3-D reconstruction method based on neural network (NN) is proposed for blood vessels skeleton from biplane angiography system and B-spline interpolation is employed to improve reconstruction performance.
- Spaceborne SAR imaging by exact compensation of the attitude errors1
A model-based algorithm is proposed to exactly compensate the error introduced by the platform attitude instability, the efficiency of the algorithm is tested with computer simulation.
- –
- –
- –
- –
- –
- Semi-Supervised High Dynamic Range Image Reconstructing via Bi-Level Uncertain Area Masking–
This work attempts to address the problem of annotation-efficient HDR image reconstructing from the view of semi-supervised learning where a teacher model generates pseudo HDR GTs for the LDR samples without GTs and a student model learns from pseudo GTs.
- Practical Test and Inference on the Inheritance of Dual Multi-Factors and Tri-Normal Distributions of Quantitative Characters–
It was proven that normal distribution of phenotypic value for quantitative traits in a segregated population was formed by adding normal distribution of genotypic value and environmental effect, which enables the genetic model to be tested in practice.
- –
- –
- –
- Pseudo-Labeling by Multi-Policy Viewfinder Network for Image Cropping–
The multi-policy viewfinder network (MPV-Net) is proposed that offers diverse refining policies to rectify the mistakes in original pseudo labels from the teacher to rectify the mistakes in original pseudo labels from the teacher.
- –
- –
- –
- –
- CaCCNN: context-aware cascaded CNN for face detection–
This paper proposes a context-aware cascaded face detector, which aggregates more information for each candidate window, and proposes an improved strategy to perform non-maximum suppression that achieves better efficiency when the number of proposals is huge.
- The Accurate Guidance for Image Caption Generation–
This work proposes the accurate guidance for image caption generation, which guides the caption model to focus more on the principle semantic object while making human reading sentence, and generate high quality sentence in grammar.
- RGB-D Co-Segmentation on Indoor Scene with Geometric Prior and Hypothesis Filtering–
This work focuses on the task of co-segmenting main furniture of indoor scene and fully utilize the complementary information of RGB-D images, and introduces a simple but effective geometric prior and two-stage hypothesis filtering strategy to refine both global and local object candidate generation.
- Stereo matching using neighboring system constructed with MST–
This work utilizes minimum spanning tree (MST) to construct a new superpixel-based neighboring system and applies this system to the stereo matching problem, which uses the superpixel as the matching primitive.
- A novel sparse-to-dense depth map generation framework for monocular videos–
A new sparse-to-dense diffusion framework is put forward that can robustly generate the depth map from monocular videos and is shown to be able to tackle the multi-view matching problem well.
- –
- mEdgeBoxes: objectness estimation for depth image–
This study proposes mEdgeboxes to evaluate the objectness in depth image with outperforms the state-of-the-art methods on the detection rate and extracts another edge map from the orientation information based on the normal vector.
- A robust texture descriptor based on a gradient orientation and modulus matrix–
Experimental results on the KTH-TIPS2 (The Royal Institute of Technology - Textures under varying Illumination, Pose and Scale) image database show that the GOMM significantly outperforms the other classical descriptors.
- Shadow detection and removal based on the saliency map–
A new method to detect the shadow in a single nature image with the saliency map and to remove the shadow is presented, which is transferred to 2D module in order to improve the blue component and compensated by enhancing the image in RGB space.
- An approach for in-harbor ship detection in complex background–
Experimental results suggest that the proposed method for ship segmentation and false alarm suppressing based on k-means clustering segmentation has high detection accuracies and low false alarm rate.
- Salient region detection Using Wasserstein distance measure based on nonlinear scale space–
The pixel level center-surround difference in a nonlinear scale space which makes blurring locally adaptive to the image regions is computed which outperforms 5 recent state of the art saliency detection methods in terms of precision and recall on a newly published benchmark.
- Localization of buildings in airborne forward-looking infrared image using template matching method–
A new approach to localize buildings from forward looking infrared (FLIR) images that can localize not only large buildings, but also small buildings is proposed and is also robust with those FLIR images degraded by clouds.
- –
- Detail-preserving image recovery based on respective restoration on phase and amplitude–
A restoration method is proposed, which cuts down phase-pollution caused by traditional regularization methods, that amplitude restoration is realized by regularization and phase restoration is achieved by algebraic method.
- A fast algorithm for wide baseline match basing on feature points filtration–
Experiments demonstrate, compared with Lowe's SIFT, that the proposed method not only decreases computing tasks, reduces consumed time, but also improves correct match ratio.
- Automatic 3D reconstruction of urban area from low resolution aerial lidar data–
An improved covariance analysis method is proposed to remove the irrelevant part of LiDAR, and a classification method based on the Markov Random Field on the connected component is proposed in this paper.
- A change detection algorithm for man-made objects based on remote sensing images–
A novel change detection algorithm is proposed to detect changes of man-made objects in remote sensing images and experimental results demonstrate the feasibility and effectiveness of the proposed algorithm.
- Rock detection based on shadows and texture for safe lunar landing–
An improved method for rock detection based on shadows and texture for safe lunar landing, which can detect and model the rocks in the lunar surface image precisely.
- An automatic DSM and remote sensing images registration scheme using template matching technique–
A new feature descriptor is presented which can highlight the similarities of DSM and remote sensing images, and can be used to form a new kind of feature image, and a criterion to choose the proper region from the feature image as the template which will ensure perfect template matching performance.
- Erratum to: A humanized anti-osteopontin antibody inhibits breast cancer growth and metastasis in vivo–
In the original publication under the Section “Results” the sentence in the Wrst paragraph for the subheading “Hu1A12 inhibits OPN-mediated anti-apoptotic and prosurvival functions” is incorrectly published as “In the presence of OPN, [3H]thymidine uptake increased 3.64-fold in MDA-MB-435S cells”.
- Analysis of matching algorithm for infrared images based on anisotropic diffusion equation–
Comparing with other preprocessing algorithms, this method achieve better performance for target recognition within low contrast and low SNR infrared images.
- The optimized parallel implementation for an improved fast correlation matching algorithm on multi-DSPs interconnected by FPGA–
An improved recursive pyramid correlation image matching algorithm is presented and the algorithm's parallel implementation on Multi-DSPs interconnected by FPGA with using several effective optimization strategies is accomplished.
- Implementation of a DSP based real-time video monitoring system–
This paper presents an implementation of a real-time video surveillance system based on TMS320DM6437 DVDP (Digital Video Development Platform), and proposes an improved motion detection method based on adaptive Gaussian Mixture Model.
- Joint distribution of nonsubsampled contourlet domain and its application to texture retrieval–
The sample covariance matrix feature is proposed and the result shows that the proposed feature is efficient in representing the texture and the difference of the joint distribution of the Nonsubsampled Contourlet Transform coefficients.
- A preprocessing method based on non-subsampled contourlet transform–
Experiments based on template matching in forward-looking infrared images show that the proposed preprocessing method based on non-subsampled contourlet transform (NSCT) can significantly improve the detection rate (DR) and the stability.
- –
- Novel framework for producing multi-scale and multi-viewpoint images based on remote sensing stereopair–
This paper proposed a framework to produce multi-scale and multi-viewpoint projection images based on remote sensing satellite stereopair by means of photogrammetry and computer vision and proposed a algorithm for reconstructing the3D structure of the target by digital photogrammetric techniques and establishing the 3D model of thetarget using the OpenGL visualization toolkit.
- Novel image matching confidence fusion evaluation algorithm based on support vector machine–
This paper firstly provides a method of how to prepare the character parameters which can accurately reflect the matching performance and then the SVM based on Gaussian kernel is used as a classifier to classify the match result and discriminate the matching result's type.
- Object recognition based on Gabor filter banks and feature decomposition modular neural networks–
A feature decomposition modular neural network classifier based on the feature extraction by the Gabor filter banks is presented for object recognition problems and the influence of band limited noise.
- Design and analysis of fuzzy controllers with amending factor–
In this paper, a fuzzy controller with amending factors is proposed and illustrated through a specific example and the control effects with respect to different number of amends factors are compared.
- Analysis of image segmentation aimed for sense matching–
A new application of the image segmentation-image matching in real-time systems is proposed and the key problem is that the best segmentation methods which are suitable to the following matching processing are provided.
- Simulation platform on scene matching–
This paper focuses on the space registration process of two images taken from the same scene by two different sensors so that their relative displacement is gotten.
- –
- –
- Expert System For Quartz Resonator Design–
Using fundamentals of artificial intelligence and object oriented technology, an expert system for quartz resonator design is developed and test results show the system can produce satisfactory design limits.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.