A5075420600- CCD and CMOS Imaging Sensors
- Advanced Memory and Neural Computing
- Gaze Tracking and Assistive Technology
- EEG and Brain-Computer Interfaces
- Neural dynamics and brain function
- Traumatic Brain Injury and Neurovascular Disturbances
- Thin-Film Transistor Technologies
- Infrared Target Detection Methodologies
- Semiconductor materials and devices
- Network Packet Processing and Optimization
- VLSI and Analog Circuit Testing
- Advanced Neural Network Applications
- Visual Attention and Saliency Detection
- Heme Oxygenase-1 and Carbon Monoxide
- Sparse and Compressive Sensing Techniques
- Transition Metal Oxide Nanomaterials
- S100 Proteins and Annexins
Tianjin Medical University General Hospital
2025
Institute of Computing Technology
2024
Washington University in St. Louis
2022-2024
Tsinghua University
2019
CMOS Image Sensors (CIS) are fundamental to emerging visual computing applications. While conventional CIS purely imaging devices for capturing images, increasingly integrate processing capabilities such as Deep Neural Network (DNN). Computational expand the architecture design space, but date no comprehensive energy model exists. This paper proposes CamJ, a detailed modeling framework that provides component-level breakdown computational and is validated against nine recent chips. We use...
With the rapid advances of deep learning-based computer vision (CV) technology, digital images are increasingly consumed, not by humans, but downstream CV algorithms. However, capturing high-fidelity and high-resolution is energy-intensive. It only dominates energy consumption sensor itself (i.e. in low-power edge devices), also contributes to significant memory burdens performance bottlenecks later storage, processing, communication stages. In this paper, we systematically explore a new...
Many computer vision tasks, ranging from recognition to multi-view registration, operate on feature representation of images rather than raw pixel intensities. However, conventional pipelines for obtaining these representations incur significant energy consumption due pixel-wise analog-to-digital (A/D) conversions and costly storage computations. In this paper, we propose HOGEye, an efficient near-pixel implementation a widely-used extraction algorithm—Histograms Oriented Gradients (HOG)....
Visual computing is vital for numerous applications. In conventional visual systems, CMOS image sensors (CIS) act as pure imaging devices capturing images, however, recent CIS designs increasingly integrate processing capabilities such Deep Neural Networks (DNN), which give rise to a notion of in-sensor computing. this paper, we propose new concept, learned computing, exploits end-to-end optimization and downstream vision tasks achieve better overall algorithm accuracy adopts...
In always-on intelligent visual perception applications, Convolution Neural Network algorithm has been widely applied. However, the implementation of it's processing is limited by hardware resources and energy budgets, which come from data access, interface overhead inflexible configuration. this paper, we present a novel analog architecture that implements low-precision Binary fuses with CMOS image sensor for applications. By enabling current-mode computation in domain, analog-to-digital...
Eye tracking is becoming an increasingly important task domain in emerging computing platforms such as Augmented/Virtual Reality (AR/VR). Today's eye system suffers from long end-to-end latency and can easily eat up half of the power budget a mobile VR device. Most existing optimization efforts exclusively focus on computation pipeline by optimizing algorithm and/or designing dedicated accelerators while largely ignoring front-end any pipeline: image sensor. This paper makes case for...