RPC: A Large-Scale Retail Product Checkout Dataset

RPC数据集是专为零售自动结账(ACO)任务设计的大型数据集，包含200个产品类别和83,739张图像。该数据集不仅规模庞大，而且在产品类别和图像数量上均居首位。数据集中的图像分为两类：一类是在受控环境中拍摄的单个产品示例图像，另一类是在结账柜台拍摄的多产品结账图像。这些图像反映了真实结账场景的复杂性，如产品随机组合、自由放置、遮挡和复杂杂乱等。此外，数据集提供了从弱到强的不同层次监督，包括购物清单、点级标注和边界框，为解决ACO问题提供了丰富的研究方向。

The RPC dataset is a large-scale dataset specifically designed for the retail automatic checkout (ACO) task, which contains 200 product categories and 83,739 images. It not only boasts a large scale but also ranks first in terms of both product categories and the number of images. The images in the dataset are divided into two categories: one is single product sample images captured in a controlled environment, and the other is multi-product checkout images taken at checkout counters. These images reflect the complexity of real-world checkout scenarios, such as random product combinations, unconstrained placement, occlusion, and complex clutter. Additionally, the dataset provides varying levels of supervision from weak to strong, including shopping lists, point-level annotations, and bounding boxes, offering abundant research directions for addressing the ACO problem.

搜集汇总

背景与挑战

背景概述

RPC数据集是专为零售自动结账任务设计的大型数据集，包含200个产品类别和超过8.3万张图像，规模领先。它提供两类图像：受控环境下的单产品示例和真实结账场景的多产品图像，后者模拟了产品随机组合、遮挡和杂乱等复杂情况。此外，数据集还提供从弱到强的多层次监督信息，如购物清单和边界框，为自动结账研究提供了丰富资源。