通过迭代缩放与细化实现精确的千兆像素人群计数
Accurate Gigapixel Crowd Counting by Iterative Zooming and Refinement
AI总结：
针对千兆像素分辨率超出GPU内存和计算极限的人群计数难题，本文提出GigaZoom方法，通过迭代放大图像密集区域并细化密度图，在千兆像素人群计数任务上取得最优性能，将次优方法准确率提升42%。
摘要：
千兆像素分辨率的日益普及给人群计数带来了新挑战。此类分辨率远超当前GPU的内存和计算极限，且现有的深度神经网络架构和训练流程并非为如此庞大的输入而设计。尽管已有若干方法被提出以应对这些挑战，但它们要么仅限于将输入图像下采样至较小尺寸，要么借鉴其他千兆像素任务，均未针对人群计数进行专门定制。本文提出了一种名为GigaZoom的新方法，该方法迭代地放大图像中最密集的区域，并用更精细的细节来细化较粗糙的密度图。实验表明，GigaZoom在千兆像素人群计数任务上取得了当前最优性能，将次优方法的准确率提升了42%。
英文摘要：
The increasing prevalence of gigapixel resolutions has presented new challenges for crowd counting. Such resolutions are far beyond the memory and computation limits of current GPUs, and available deep neural network architectures and training procedures are not designed for such massive inputs. Although several methods have been proposed to address these challenges, they are either limited to downsampling the input image to a small size, or borrowing from other gigapixel tasks, which are not tailored for crowd counting. In this paper, we propose a novel method called GigaZoom, which iteratively zooms into the densest areas of the image and refines coarser density maps with finer details. Through experiments, we show that GigaZoom obtains the state-of-the-art for gigapixel crowd counting and improves the accuracy of the next best method by 42%.
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。
