用于推理的硬连线分布式处理单元:在传感器端进行神经决策以释放主机资源
Hardwired Distributed Processing Units for Inference: Neural Decisions at the Sensor that Free the Host
AI总结:
本研究提出硬连线分布式处理单元(HDPU),将推理决策从主机迁移至传感器端,在Artix-7 FPGA上实现斑马线检测,准确率达95.7%,大幅降低主机计算与数据传输负载。
摘要:
我们证明了硬连线分布式处理单元(HDPU)可在传感器端做出神经决策,从而将推理任务从主机中移除。该单元基于低成本Artix-7 FPGA构建,用于斑马线检测,在固定的2269477个时钟周期(34.04 ms,取决于FPGA时钟)内对每帧相机图像做出决策,且该时间处于37.49 ms的帧周期内;在所有639张测试图像上,其返回的分数和决策与比特精确的软件模型完全一致。从相机像素到决策的整个检测路径仅占用1814个查找表(LUT),无需处理器、操作系统或全分辨率帧存储器。当该单元不回答“不确定”而直接给出结果时,其在街景照片上的回答准确率达95.7%;硬件对比度阶段将相机式行车记录仪帧的准确率从47.7%提升至78.2%。主机无需承担任何推理任务:每帧周期内可节省8.13 ms的核心时间,而全精度ResNet-18即使使用16线程也需要超过一个帧周期才能完成推理;主机接收的每帧数据量仅为16字节,远低于原本的614400字节,压缩了38400倍。该单元是一个80×60像素的流水线,为四层二值化卷积神经网络(CNN)提供数据,该网络包含93408个1比特权重,已通过Python模型到RTL仿真再到硬件板的比特精确验证。早期HDPU仅执行预处理(边缘、角点、转向),而本研究表明决策本身可迁移至传感器端。
英文摘要:
We show that a hardwired distributed processing unit (HDPU) can take a neural decision at the sensor and so remove the inference from the host. Built for zebra-crossing detection on a low-cost Artix-7 FPGA, the unit decides on every camera frame in a fixed 2,269,477 clock cycles (34.04 ms depends upon Fpga clk ), inside the 37.49-ms frame period, and on the board it returns the same scores and decisions as its bit-exact software model on all 639 test images. The whole detection path, from camera pixels to decision, takes 1,814 LUTs, with no processor, no operating system and no full-resolution frame store. When the unit answers instead of reporting "not sure", 95.7% of its answers on street photographs are correct, and a hardware contrast stage lifts accuracy on camera-like dashcam frames from 47.7% to 78.2%. The host is spared all of the inference: 8.13 ms of a core in every frame period, where a full-precision ResNet-18 would need more than a frame period even on 16 threads. It also receives 16 bytes per frame instead of 614,400, 38,400 times less. The unit is an 80x60-pixel pipeline feeding a four-layer binarized CNN with 93,408 one-bit weights, verified bit-exactly from a Python model through RTL simulation to the board. Earlier HDPUs did pre-processing (edges, corners, steering); this one shows that the decision itself can move to the sensor.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。