A5016814203- Particle Detector Development and Performance
- Advanced Memory and Neural Computing
- Particle physics theoretical and experimental studies
- CCD and CMOS Imaging Sensors
- Neutrino Physics Research
- Advanced Neural Network Applications
- Satellite Image Processing and Photogrammetry
- Radiation Effects in Electronics
- Robotic Path Planning Algorithms
- Robotics and Sensor-Based Localization
- Astrophysics and Cosmic Phenomena
- Optical measurement and interference techniques
- Real-Time Systems Scheduling
- Embedded Systems Design Techniques
- Video Coding and Compression Technologies
- Advanced Image Processing Techniques
- Distributed and Parallel Computing Systems
- Electrostatic Discharge in Electronics
- Ferroelectric and Negative Capacitance Devices
- Interconnection Networks and Systems
- Advanced Data Storage Technologies
- Radiation Detection and Scintillator Technologies
- Image and Video Quality Assessment
- Remote Sensing and LiDAR Applications
- Advanced Vision and Imaging
META Health
2024
University of Michigan
2017-2022
Meta (United States)
2020
The use of Light Detection And Ranging (LiDAR) has enabled the continued improvement in accuracy and performance autonomous navigation. latest applications require LiDAR's highest spatial resolution, which generate a massive amount 3D point clouds that need to be processed real time. In this work, we investigate architecture design for k-Nearest Neighbor (kNN) search, an important processing kernel clouds. An approximate kNN search based on k-dimensional (k-d) tree is employed improve...
Machine learning algorithms have enabled high quality stereo depth estimation to run on Augmented and Virtual Reality (AR/VR) devices. However, energy consumption across the full image processing stack prevents from running effectively battery-limited This paper introduces SteROI-D, a system paired with mapping methodology. SteROI-D exploits Region-of-Interest (ROI) temporal sparsity at level save energy. SteROI-D's flexible heterogeneous compute fabric supports diverse ROIs. Importantly, we...
Untethered Augmented and Virtual Reality (AR/VR) devices are an emerging compute platform with unique opportunities challenges. AR/VR use array of sensors, including multiple cameras, to understand their surroundings provide the user immersive experience. To deliver functionality performance, rely on state-of-the-art algorithms Deep Neural Networks (DNNs). These must operate in real time, it presents a computational challenge for mobile system. The emergence on-sensor provides possible...
Augmented Reality/Virtual Reality (AR/VR) glasses are widely foreseen as the next generation computing platform. AR/VR a complex "system of systems" which must satisfy stringent form factor, computing-, power- and thermal- requirements. In this paper, we will show that novel distributed on-sensor compute architecture, coupled with new semiconductor technologies (such dense 3D-IC interconnects Spin-Transfer Torque Magneto Random Access Memory, STT-MRAM) and, most importantly, full...
We present an implementation of fixed-latency gigabit serial links in a low-cost Xilinx field-programmable gate array. The is targeted for data packet router the upgrade ATLAS muon spectrometer. serves as switch. It handles up to 12 inputs at 4.8 Gbps from on-detector electronics and four 4.8-Gbps outputs trigger processing circuits. input streams are deserialized aligned common clock domain NULL suppression forwarding. Gigabit transceivers used processing, scheme developed maintain low...
We present TAICHI, a general in-memory computing deep neural network accelerator design based on RRAM crossbar arrays heterogeneously integrated with local arithmetic units and global co-processors to allow the system efficiently map different models while maintaining high energy efficiency throughput. A hierarchical mesh network-on-chip is implemented facilitate communication among clusters in TAICHI balance reconfigurability efficiency. Detailed deployment of circuit components discussed,...
In wearable AR/VR systems, data transmission between cameras and central processors can account for a significant portion of total system power, particularly in high framerate applications. Thus, it becomes necessary to compress video streams reduce the cost transmission. this paper we present CNN-based compression scheme such vision systems. We demonstrate that, unlike conventional techniques, our method be tuned specific machine This enables increased given application performance target....
Advanced edge sensing/computing devices, such as AR/VR have a uniquely challenging adaptive baseline workload and camera sensor structure. These devices must process images in real-time from multiple sensors, placing large burden on typical centralized mobile SoC processor. Augmenting the sensors with package-integrated near-sensor processor can improve device's processing performance well reduce energy consumption. This adapt to dynamic workloads, fit within limited silicon footprint...