Indoor Object Detection Dataset

The indoor object detection dataset is divided into three parts: the training set (94%), validation set (4%), and test set (2%), with 12,012 images for training, 490 for validation, and 245 for testing. The dataset consists of seven classes, which are chair, door, human, table, bin, fire extinguisher, and shelf. These include 1,985 instances of chairs, 1,735 instances of doors, 1,011 instances of humans, 809 instances of tables, 696 instances of bins, 534 instances of fire extinguishers, and 288 instances of shelves. Moreover, the training set of the dataset contains 1,299 instances of chairs, 1,150 instances of doors, 673 instances of humans, 533 instances of tables, 469 instances of bins, 364 instances of fire extinguishers, and 188 instances of shelves. Additionally, the validation set of the dataset contains 429 instances of chairs, 305 instances of doors, 216 instances of humans, 161 instances of tables, 135 instances of bins, 97 instances of fire extinguishers, and 48 instances of shelves. Furthermore, the test set of the dataset contains several classes of everyday objects, with the following distribution: 175 instances of doors, 168 instances of chairs, 83 instances of humans, 67 instances of tables, 65 instances of bins, 49 instances of fire extinguishers, and 20 instances of shelves. In addition, preprocessing steps included auto-orientation and resizing all images to 640×640 pixels to maintain uniformity. To improve generalization for real-world applications, we applied data augmentation techniques, including horizontal and vertical flipping, 90-degree rotations (clockwise, counterclockwise, and upside down), random rotations within -15° to +15°, shearing within ±10° horizontally and vertically, and brightness adjustments between -15% and +15%. Additionally, this annotated, preprocessed, and augmented dataset enhances object detection accuracy in indoor scenes.

本室内目标检测数据集划分为训练集（94%）、验证集（4%）与测试集（2%）三个子集，其中训练集包含12012张图像，验证集490张，测试集245张。该数据集共涵盖7个目标类别，分别为椅子（chair）、门（door）、行人（human）、桌子（table）、垃圾桶（bin）、灭火器（fire extinguisher）及货架（shelf）。各类别总实例数如下：椅子1985个、门1735个、行人1011个、桌子809个、垃圾桶696个、灭火器534个、货架288个。训练集的各类别实例数为：椅子1299个、门1150个、行人673个、桌子533个、垃圾桶469个、灭火器364个、货架188个。验证集的各类别实例数为：椅子429个、门305个、行人216个、桌子161个、垃圾桶135个、灭火器97个、货架48个。测试集的各类别日常物品实例分布为：门175个、椅子168个、行人83个、桌子67个、垃圾桶65个、灭火器49个、货架20个。数据集预处理环节包含自动图像方向校正与统一尺寸调整，将所有图像缩放至640×640像素以保证数据一致性。为提升模型在实际应用场景中的泛化能力，我们采用了多种数据增强技术，包括水平与垂直翻转、90度旋转（顺时针、逆时针及上下翻转式旋转）、-15°至+15°范围内的随机旋转、±10°范围内的水平与垂直剪切变换，以及亮度调整幅度为-15%至+15%的亮度修正。该经过标注、预处理与数据增强的数据集，可有效提升室内场景下的目标检测精度。

搜集汇总

背景与挑战

背景概述

该数据集是一个室内目标检测数据集，包含12,747张图像，分为7个常见物体类别，所有图像经过标准化预处理并应用了多种数据增强技术。数据集已划分为训练集(94%)、验证集(4%)和测试集(2%)，适用于开发室内环境下的目标检测系统。