A 65nm 24.7µJ/Frame 12.3mW Activation-SimilarityAware Convolutional Neural Network Video Processor

📋 论文概要

本文提出了一种激活相似性感知的卷积神经网络视频处理器，通过混合精度计算、帧间数据重用和混合位宽差分技术，显著降低了计算和存储功耗。在65nm工艺下实现了12.3mW功耗和24.7µJ/帧的效率，适用于自动驾驶和监控等视频应用。

💡 主要创新点

• 提出激活相似性感知架构，利用帧间激活值的相似性跳过冗余计算，降低功耗。
• 采用混合精度和混合位宽差分技术，灵活调整计算精度以平衡功耗与精度。
• 实现帧间数据重用，减少外部存储器访问，进一步降低数据搬运能耗。

🏷 关键词

📄 原文摘要

Zhe Yuan1,2, Yixiong Yang1, Jinshan Yue1,2, Ruoyang Liu1, Xiaoyu Feng1, Zhiting Lin3, Xiulong Wu3, Xueqing Li1, Huazhong Yang1, Yongpan Liu1 Tsinghua University, Beijing, China Pi2star Technology, Beijing, China 3 Anhui University, Hefei, China 1 2 Convolutional Neural Networks (CNNs) have become widely used in image signal processing, such as tracking, classification and post-processing. Modern CNNs use millions of weights and activations, leading to critical challenges for both computation and data transmission. Video applications, such as autopilot and surveillance cameras, have to process a large number of sequential images/frames within limited time, making the situation even worse. As shown in Fig. 14.2.1, adjacent activation frames of typical video applications are similar to each other most of the time, providing an opportunity to reduce both computing and data transmission complexity significantly.

👥 作者与机构

Using Hybrid Precision, Inter-Frame Data Reuse and, Mixed-Bit-Width Difference-Frame Data Codec