A Sparse Convolution Neural Network Accelerator for 3D/4D Point-Cloud Image Recognition on Low Power Mobile Device with Hopping-Index Rule Book for Efficient Coordinate Management
Qiankai Cao, Jie Gu
20222022 IEEE Symposium on VLSI Technology and Circuits (VLSI Technology and Circuits)27 citationsDOI
Abstract
This work presents the first 3D/4D sparse CNN (SCNN) accelerator for point cloud image recognition on low power devices. A special hopping-index rule book method and efficient data search technique were developed to mitigate the overhead of coordinate management for SCNN. A 65nm test chip for 3D/4D images was demonstrated with 7.09–13.6 TOPS/W power efficiency and state-of-the-art frame rate.
Topics & Concepts
Computer scienceConvolution (computer science)Convolutional neural networkPoint cloudCloud computingOverhead (engineering)Artificial intelligencePower (physics)Index (typography)Frame (networking)Image (mathematics)ChipPoint (geometry)Artificial neural networkComputer engineeringComputer visionComputer hardwareOperating systemTelecommunicationsMathematicsQuantum mechanicsWorld Wide WebPhysicsGeometryOptical measurement and interference techniquesAdvanced Optical Sensing TechnologiesCCD and CMOS Imaging Sensors