Deep learning

Applications

Point cloud segmentation

• RangeSeg: Range-Aware Real Time Segmentation of 3D LiDAR Point Clouds

IMU based gait phase and related estimation

• IMU Based Deep Stride Length Estimation With Self-Supervised Learning

• Real-time wearable gait phase segmentation for running and walking

• Deep gait tracking with inertial measurement unit

• Real Time On Sensor Gait Phase Detection with 0.5 KB Deep Learning Model

• Determining motions with an IMU during level walking and slope and stair walking

Segmentation for medical images

• Semantic segmentation of intracranial hemorrhages in head CT scans

Model accelerations

Reduce memory bandwidth

• Zebra: Memory Bandwidth Reduction for CNN Accelerators with Zero Block Regularization of Activation Maps

• Run time adaptive network slimming for mobile environments

Hardware accelerations (Digital implementation)

Spiking neural network

• Sparse Compressed Spiking Neural Network Accelerator for Object Detection

Convolutional neural network

• A Real-Time 1280, x, 720 Object Detection Chip With 585 MB/s Memory Traffic

• Pre-RTL DNN Hardware Evaluator With Fused Layer Support

• VSA: Reconfigurable vectorwise spiking neural network accelerator

• Efficient accelerator for dilated and transposed convolution with decomposition

• VWA: Hardware efficient vectorwise accelerator for convolutional neural network

• VSCNN: convolution neural network accelerator with vector sparsity

• End-to-end hardware accelerator for deep convolutional neural network

• Data and hardware efficient design for convolutional neural network

Hardware accelerations (In-memory computing implementation)

• Device quantization policy in variation-aware in-memory computing design

• A 48 TOPS and 20943 TOPS/W 512kb Computation-in-SRAM Macro for Highly Reconfigurable Ternary CNN Acceleration

• Device Quantization Policy and Power-Performance-Area Co-Optimization Strategy in Variation-Aware In-memory Computing Design

• Sub-nA low-current HZO ferroelectric tunnel junction for high-performance and accurate deep learning acceleration

• NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive AI edge

• Challenges and opportunities toward online training acceleration using RRAM-based hardware neural network

• Mitigating asymmetric nonlinear weight update effects in hardware neural network based on analog resistive synapse