A5032271514- Visual Attention and Saliency Detection
- Advanced Neural Network Applications
- Face Recognition and Perception
- Advanced Algorithms and Applications
- Brain Tumor Detection and Classification
- Advanced Memory and Neural Computing
- Image and Video Quality Assessment
- Reinforcement Learning in Robotics
- Distributed and Parallel Computing Systems
- CCD and CMOS Imaging Sensors
- Advanced Image and Video Retrieval Techniques
- Domain Adaptation and Few-Shot Learning
- Data Quality and Management
- Virtual Reality Applications and Impacts
- Infrared Target Detection Methodologies
- Hydraulic and Pneumatic Systems
- Gas Sensing Nanomaterials and Sensors
- Advanced Chemical Sensor Technologies
- Analytical Chemistry and Sensors
- Inertial Sensor and Navigation
- Diverse Musicological Studies
- Robotics and Automated Systems
- Smart Grid Security and Resilience
- Embedded Systems and FPGA Design
- Gaze Tracking and Assistive Technology
Harbin Institute of Technology
2019-2025
Heilongjiang Institute of Technology
2023
Beihang University
2018
The Convolutional Neural Network (CNN) has been used in many fields and achieved remarkable results, such as image classification, face detection, speech recognition. Compared to GPU (graphics processing unit) ASIC, a FPGA (field programmable gate array)-based CNN accelerator great advantages due its low power consumption reconfigurable property. However, FPGA’s extremely limited resources CNN’s huge amount of parameters computational complexity pose challenges the design. Based on ZYNQ...
Recent advances in deep convolution neural networks (CNNs) boost the development of video salient object detection (SOD), and many remarkable deep-CNNs SOD models have been proposed. However, existing still suffer from coarse boundaries object, which may be attributed to loss high-frequency information. The traditional graph-based can preserve well by conducting superpixels/supervoxels segmentation advance, but they perform weaker highlighting whole than latest models, limited heuristic...
Convolutional neural network (CNN)-based salient object detection (SOD) models have achieved promising performance in optical remote sensing images (ORSIs) recent years. However, the restriction concerning local sliding window operation of CNN has caused many existing CNN-based ORSI SOD to still struggle with learning long-range relationships. To this end, a novel transformer framework is proposed for SOD, which inspired by powerful global dependency relationships networks. This first...
We present a novel, robust estimation method to distinguish salient objects from complicated, dynamic backgrounds in videos. In this method, we propose novel approach model motion energy based on magnitude, orientation, gradient flow field, and spatial of the video frame. Furthermore, an effective spatiotemporal objectness map is also proposed estimate compact object-like region current frame leveraging both proposals saliency previous Then oversegmented into granularity superpixels using...
Continual learning is a subfield of machine learning, which aims to allow models continuously learn on new data, by accumulating knowledge without forgetting what was learned in the past. In this work, we take step back, and ask: "Why should one care about continual first place?". We set stage examining recent papers published at four major conferences, show that memory-constrained settings dominate field. Then, discuss five open problems even though they might seem unrelated sight, will...
Multi-agent reinforcement learning (MARL) has become more and popular over recent decades, the need for high-level cooperation is increasing every day because of complexity real-world environment. However, multi-agent credit assignment problem that serves as main obstacle to coordination still not addressed properly. Though lots methods have been proposed, none them thought perform assignments across multi-levels. In this paper, we aim propose an approach a better scheme by First,...
Convolution neural networks (CNN), support vector machine (SVM) and hybrid CNN-SVM algorithms are widely applied in many fields, including image processing fault diagnosis. Although dedicated FPGA accelerators have been proposed for specific networks, such as CNN or SVM, few of them focused on CNN-SVM. Furthermore, the existing do not CNN-SVM, which limits their application scenarios. In this work, we propose a accelerator FPGA. This utilizes novel hardware-reuse architecture unique...
With the development of Internet Things (IoT) and edge computing technology, gas sensor arrays based on Micro-Electro-Mechanical System (MEMS) fabrication technique have broad application prospects in intelligent integrated systems, portable devices, other fields. In such complex scenarios, normal operation a sensing system depends heavily accuracy output. Therefore, lightweight Self-Detection Self-Calibration strategy for MEMS is proposed this paper to monitor working status correct...
Self-powered wearable sweat-lactate analyzer has been developed for training analysis of rowing. Tetrapodshaped ZnO nanowires are attached onto ordinary textiles to form a device. Based on the coupling enzymatic reaction (lactate oxidase and lactic acid) piezoelectric effect, device can be tester monitor driving frequency, rowing distance sweat acid concentration in real-time. The relationship between frequency physiological state is obtained by monitoring process tester, which helpful...
Matrix multiplication is a critical time-consuming processing step in many machine learning applications. Due to the diversity of practical applications, matrix dimensions are generally not fixed. However, most calculation methods, based on field programmable gate array (FPGA) currently use fixed dimensions, which limit flexibility algorithms FPGA. The bottleneck lies limited FPGA resources. Therefore, this paper proposes an accelerator architecture for computing method with changeable...
Automatic music transcription (AMT) aims to convert raw audio signals into symbolic music. This is a highly challenging task in the fields of signal processing and artificial intelligence, it holds significant application value information retrieval (MIR). Existing methods based on convolutional neural networks (CNNs) often fall short capturing time-frequency characteristics tend overlook interdependencies between notes when polyphonic piano with multiple simultaneous notes. To address these...
Online training of Support Vector Regression (SVR) in the field machine learning is a computationally complex algorithm. Due to need for multiple iterative processing training, SVR usually implemented on computer, and existing methods cannot be directly Field-Programmable Gate Array (FPGA), which restricts application range. This paper reconstructs framework implementation without precision loss reduce total latency required matrix update, reducing time consumption by 90%. A general ε-SVR...