论文阅读 [TPAMI-2022] Grid Anchor Based Image Cropping: A New Benchmark and An Efficient Model

论文阅读 [TPAMI-2022] Grid Anchor Based Image Cropping: A New Benchmark and An Efficient Model

论文搜索(studyai.com)

搜索论文: Grid Anchor Based Image Cropping: A New Benchmark and An Efficient Model

搜索论文: http://www.studyai.com/search/whole-site/?q=Grid+Anchor+Based+Image+Cropping:+A+New+Benchmark+and+An+Efficient+Model

关键字(Keywords)

Agriculture; Measurement; Databases; Robustness; Benchmark testing; Training; Image cropping; photo cropping; image aesthetics; deep learning

机器视觉

数据扩增

摘要(Abstract)

Image cropping aims to improve the composition as well as aesthetic quality of an image by removing extraneous content from it.

图像裁剪旨在通过去除图像中的无关内容来改善图像的构图和审美质量。.

Most of the existing image cropping databases provide only one or several human-annotated bounding boxes as the groundtruths, which can hardly reflect the non-uniqueness and flexibility of image cropping in practice.

现有的图像裁剪数据库大多只提供一个或多个人类标注的边界框作为基本事实，这很难反映实际中图像裁剪的非唯一性和灵活性。.

The employed evaluation metrics such as intersection-over-union cannot reliably reflect the real performance of a cropping model, either.

所采用的评估指标（如联合上的交叉）也不能可靠地反映裁剪模型的真实性能。.

This work revisits the problem of image cropping, and presents a grid anchor based formulation by considering the special properties and requirements (e.g., local redundancy, content preservation, aspect ratio) of image cropping.

这项工作重新审视了图像裁剪问题，并通过考虑图像裁剪的特殊属性和要求（例如，局部冗余、内容保留、纵横比），提出了一种基于网格锚的公式。.

Our formulation reduces the searching space of candidate crops from millions to no more than ninety.

我们的公式将候选作物的搜索空间从数百万减少到不超过90。.

Consequently, a grid anchor based cropping benchmark is constructed, where all crops of each image are annotated and more reliable evaluation metrics are defined.

因此，构建了一个基于网格锚的裁剪基准，其中每个图像的所有裁剪都被注释，并定义了更可靠的评估指标。.

To meet the practical demands of robust performance and high efficiency, we also design an effective and lightweight cropping model.

为了满足高性能和高效率的实际需求，我们还设计了一种高效、轻量级的裁剪模型。.

By simultaneously considering the region of interest and region of discard, and leveraging multi-scale information, our model can robustly output visually pleasing crops for images of different scenes.

通过同时考虑感兴趣区域和丢弃区域，并利用多尺度信息，我们的模型可以为不同场景的图像稳健地输出视觉上令人愉悦的作物。.

With less than 2.5M parameters, our model runs at a speed of 200 FPS on one single GTX 1080Ti GPU and 12 FPS on one i7-6800K CPU.

在不到2.5M的参数下，我们的模型在一个GTX 1080Ti GPU上以200 FPS的速度运行，在一个i7-6800K CPU上以12 FPS的速度运行。.

The code is available at: https://github.com/HuiZeng/Grid-Anchor-based-Image-Cropping-Pytorch…

该代码可从以下网址获取：https://github.com/HuiZeng/Grid-Anchor-based-Image-Cropping-Pytorch…

作者(Authors)

[‘Hui Zeng’, ‘Lida Li’, ‘Zisheng Cao’, ‘Lei Zhang’]