A5006986293- Video Surveillance and Tracking Methods
- Visual Attention and Saliency Detection
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Advanced Vision and Imaging
- Human Pose and Action Recognition
- Face recognition and analysis
- Multimodal Machine Learning Applications
- Image Enhancement Techniques
- Advanced Image Processing Techniques
- Olfactory and Sensory Function Studies
- Advanced Image Fusion Techniques
- Face Recognition and Perception
- Domain Adaptation and Few-Shot Learning
- Image and Video Quality Assessment
- Face and Expression Recognition
- Infrared Target Detection Methodologies
- Gaze Tracking and Assistive Technology
- Image Processing Techniques and Applications
- Remote-Sensing Image Classification
- Image Retrieval and Classification Techniques
- Gait Recognition and Analysis
- Video Analysis and Summarization
- Image and Signal Denoising Methods
- Medical Image Segmentation Techniques
Dalian University of Technology
2016-2025
Dalian University
2005-2024
Peng Cheng Laboratory
2019-2023
Most existing bottom-up methods measure the foreground saliency of a pixel or region based on its contrast within local context entire image, whereas few focus segmenting out background regions and thereby salient objects. Instead considering between objects their surrounding regions, we consider both cues in different way. We rank similarity image elements (pixels regions) with via graph-based manifold ranking. The is defined relevances to given seeds queries. represent as close-loop graph...
Model distillation is an effective and widely used technique to transfer knowledge from a teacher student network. The typical application powerful large network or ensemble small network, in order meet the low-memory fast execution requirements. In this paper, we present deep mutual learning (DML) strategy. Different one-way between static pre-defined model distillation, with DML, of students learn collaboratively teach each other throughout training process. Our experiments show that...
Sparse representation has been applied to visual tracking by finding the best candidate with minimal reconstruction error using target templates. However most sparse based trackers only consider holistic and do not make full use of coefficients discriminate between background, hence may fail more possibility when there is similar object or occlusion in scene. In this paper we develop a simple yet robust method on structural local appearance model. This exploits both partial information...
Deep Neural Networks (DNNs) have substantially improved the state-of-the-art in salient object detection. However, training DNNs requires costly pixel-level annotations. In this paper, we leverage observation that image-level tags provide important cues of foreground objects, and develop a weakly supervised learning method for saliency detection using only. The Foreground Inference Network (FIN) is introduced challenging task. first stage our method, FIN jointly trained with fully...
We propose a new approach for general object tracking with fully convolutional neural network. Instead of treating network (CNN) as black-box feature extractor, we conduct in-depth study on the properties CNN features offline pre-trained massive image data and classification task ImageNet. The discoveries motivate design our system. It is found that layers in different levels characterize target from perspectives. A top layer encodes more semantic serves category detector, while lower...
In this paper we propose a robust object tracking algorithm using collaborative model. As the main challenge for is to account drastic appearance change, model that exploits both holistic templates and local representations. We develop sparsity-based discriminative classifier (SD-C) generative (SGM). S-DC module, introduce an effective method compute confidence value assigns more weights foreground than background. SGM novel histogram-based takes spatial information of each patch into...
Correlation acts as a critical role in the tracking field, especially recent popular Siamese-based trackers. The correlation operation is simple fusion manner to consider similarity between template and search region. However, itself local linear matching process, leading lose semantic information fall into optimum easily, which may be bottleneck of designing high-accuracy algorithms. Is there any better feature method than correlation? To address this issue, inspired by Transformer, work...
Fully convolutional neural networks (FCNs) have shown outstanding performance in many dense labeling problems. One key pillar of these successes is mining relevant information from features layers. However, how to better aggregate multi-level feature maps for salient object detection underexplored. In this work, we present Amulet, a generic aggregating framework detection. Our first integrates into multiple resolutions, which simultaneously incorporate coarse semantics and fine details. Then...
In this paper, we propose a visual saliency detection algorithm from the perspective of reconstruction errors. The image boundaries are first extracted via super pixels as likely cues for background templates, which dense and sparse appearance models constructed. For each region, compute Second, errors propagated based on contexts obtained K-means clustering. Third, pixel-level is computed by an integration multi-scale refined object-biased Gaussian model. We apply Bayes formula to integrate...
This paper presents a saliency detection algorithm by integrating both local estimation and global search. In the stage, we detect using deep neural network (DNN-L) which learns patch features to determine value of each pixel. The estimated maps are further refined exploring high level object concepts. search map together with contrast geometric information used as describe set candidate regions. Another (DNN-G) is trained predict score region based on features. final generated weighted sum...
In this paper, we present a new tracking architecture with an encoder-decoder transformer as the key component. The encoder models global spatio-temporal feature dependencies between target objects and search regions, while decoder learns query embedding to predict spatial positions of objects. Our method casts object direct bounding box prediction problem, without using any proposals or predefined anchors. With transformer, just uses simple fully-convolutional network, which estimates...
Deep-learning based salient object detection methods achieve great progress. However, the variable scale and unknown category of objects are challenges all time. These closely related to utilization multi-level multi-scale features. In this paper, we propose aggregate interaction modules integrate features from adjacent levels, in which less noise is introduced because only using small up-/down-sampling rates. To obtain more efficient integrated features, self-interaction embedded each...
Effective convolutional features play an important role in saliency estimation but how to learn powerful for is still a challenging task. FCN-based methods directly apply multi-level without distinction, which leads sub-optimal results due the distraction from redundant details. In this paper, we propose novel attention guided network selectively integrates contextual information progressive manner. Attentive generated by our can alleviate of background thus achieve better performance. On...
In this paper, we formulate saliency detection via absorbing Markov chain on an image graph model. We jointly consider the appearance divergence and spatial distribution of salient objects background. The virtual boundary nodes are chosen as in a absorbed time from each transient node to is computed. measures its global similarity with all nodes, thus can be consistently separated background when used metric. Since relies weights path their distance, region center may salient. further...
Deep convolutional neural networks (CNNs) have delivered superior performance in many computer vision tasks. In this paper, we propose a novel deep fully network model for accurate salient object detection. The key contribution of work is to learn uncertain features (UCF), which encourage the robustness and accuracy saliency We achieve via introducing reformulated dropout (R-dropout) after specific layers construct an ensemble internal feature units. addition, effective hybrid upsampling...
While numerous algorithms have been proposed for object tracking with demonstrated success, it remains a challenging problem tracker to handle large change in scale, motion, shape deformation occlusion. One of the main reasons is lack effective image representation account appearance variation. Most trackers use high-level structure or low-level cues representing and matching target objects. In this paper, we propose method from perspective mid-level vision structural information captured...
Recent progress on salient object detection is beneficial from Fully Convolutional Neural Network (FCN). The saliency cues contained in multi-level convolutional features are complementary for detecting objects. How to integrate becomes an open problem detection. In this paper, we propose a novel bi-directional message passing model At first, adopt Multi-scale Context-aware Feature Extraction Module (MCFEM) feature maps capture rich context information. Then structure designed pass messages...
In this paper, we introduce Cellular Automata-a dynamic evolution model to intuitively detect the salient object. First, construct a background-based map using color and space contrast with clustered boundary seeds. Then, novel propagation mechanism dependent on Automata is proposed exploit intrinsic relevance of similar regions through interactions neighbors. Impact factor matrix coherence are constructed balance influential power towards each cell's next state. The saliency values all...
The Visual Object Tracking challenge VOT2017 is the fifth annual tracker benchmarking activity organized by VOT initiative. Results of 51 trackers are presented; many state-of-the-art published at major computer vision conferences or journals in recent years. evaluation included standard and other popular methodologies a new "real-time" experiment simulating situation where processes images as if provided continuously running sensor. Performance tested typically far exceeds baselines. source...
Recent deep learning based salient object detection methods achieve gratifying performance built upon Fully Convolutional Neural Networks (FCNs). However, most of them have suffered from the boundary challenge. The state-of-the-art employ feature aggregation tech- nique and can precisely find out wherein object, but they often fail to segment entire with fine boundaries, especially those raised narrow stripes. So there is still a large room for improvement over FCN models. In this paper, we...
Effective integration of contextual information is crucial for salient object detection. To achieve this, most existing methods based on 'skip' architecture mainly focus how to integrate hierarchical features Convolutional Neural Networks (CNNs). They simply apply concatenation or element-wise operation incorporate high-level semantic cues and low-level detailed information. However, this can degrade the quality predictions because cluttered noisy also be passed through. address problem, we...