SKU110K

Sku110k 数据集提供了 11,762 张图像，在密集场景中捕获了超过 170 万个带注释的边界框，其中包括 8,233 张用于训练的图像、588 张用于验证的图像和 2,941 张用于测试的图像。总共有大约 1,733,678 个实例。这些图像是从数千家超市商店收集的，具有各种规模、视角、照明条件和噪音水平。所有图像都被调整为一百万像素的分辨率。数据集中的大多数实例都是紧密排列的，并且通常具有 [−15∘, 15∘] 范围内的特定方向。

The Sku110k dataset includes 11,762 images with over 1.73 million annotated bounding boxes captured in dense supermarket scenes. It is divided into three standard subsets: 8,233 training images, 588 validation images, and 2,941 test images, resulting in a total of approximately 1,733,678 annotated instances. All images were collected from thousands of physical supermarket stores, featuring diverse scales, shooting perspectives, lighting conditions, and noise levels. Each image is resized to a resolution of one million pixels. Most instances in the dataset are tightly arranged, and typically have orientations within the range of [-15°, 15°].

AI搜集汇总

背景与挑战

背景概述

SKU110K是一个用于密集场景目标检测的数据集，包含11,762张超市商品图像，总计约170万个带注释的边界框，覆盖多样化条件如视角、照明和噪音。该数据集专为处理紧密排列、方向特定的实例而设计，适用于计算机视觉任务如商品识别和库存管理。