arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.07598cs.CV
收藏

Mondrian:基于压缩打包推理的设备端高性能视频分析

Mondrian: On-Device High-Performance Video Analytics with Compressive Packed Inference

Seoul National University(首尔大学);Nokia Bell Labs(诺基亚贝尔实验室)

Changmin Jeon, Seonjun Kim, Juheon Yi, Youngki Lee

AI总结:

研究针对资源受限设备在高分辨率视频分析上的不足,提出压缩打包推理方法,通过选择性处理像素、动态缩小ROI并智能组合,提升目标检测性能,比现有基线准确率更高、吞吐量更大。

摘要:

本文介绍了Mondrian,一种能对高分辨率视频流进行高性能目标检测的边缘系统。许多针对资源受限设备的轻量级模型和系统优化技术未充分利用加速器在动态高分辨率视频上的潜力。为此,我们设计了新颖的压缩打包推理,通过选择性确定处理像素并组合以最大化处理并行性来最小化逐像素处理成本。系统快速提取并动态缩小感兴趣区域(ROI),然后智能组合缩放后的ROI以最大化GPU等推理加速器的利用率。跨数据集、模型和设备的评估表明,Mondrian比现有基线(如输入缩放、ROI提取、ROI提取+批处理)的准确率高15.0 - 19.7%,处理1080p视频流的吞吐量比逐帧推理高6.65倍。代码可通过该https链接获取。

英文摘要:

In this paper, we present Mondrian, an edge system that enables high-performance object detection on high-resolution video streams. Many lightweight models and system optimization techniques have been proposed for resource-constrained devices, but they do not fully utilize the potential of the accelerators over dynamic, high-resolution videos. To enable such capability, we devise a novel Compressive Packed Inference to minimize per-pixel processing costs by selectively determining the necessary pixels to process and combining them to maximize processing parallelism. In particular, our system quickly extracts ROIs and dynamically shrinks them, reflecting the effect of the fast-changing characteristics of objects and scenes. It then intelligently combines such scaled ROIs into large canvases to maximize the utilization of inference accelerators such as GPU. Evaluation across various datasets, models, and devices shows Mondrian outperforms state-of-the-art baselines (e.g., input rescaling, ROI extractions, ROI extractions+batching) by 15.0-19.7% higher accuracy, leading to $\times$6.65 higher throughput than frame-wise inference for processing various 1080p video streams. The code is available at https://github.com/snuhcs/mondrian.

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。

↑