Dataset-I-drinking-related-object-detection (in both YoloV8 and COCO format)

This dataset contains annotated images for object detection for containers and hands in a first-person view (egocentric view) during drinking activities. Both YOLOV8 format and COCO format are provided.<br>Please refer to our paper for more details.<br><b>Purpose:</b> Training and testing the object detection model.<b>Content:</b> Videos from Session 1 of Subjects 1-20.<b>Images:</b> Extracted from the videos of Subjects 1-20 Session 1.<b>Additional Images:</b>~500 hand/container images from Roboflow Open Source data.~1500 null (background) images from VOC Dataset and MIT Indoor Scene Recognition Dataset:1000 indoor scenes from 'MIT Indoor Scene Recognition'400 other unrelated objects from VOC Dataset<b>Data Augmentation:</b>Horizontal flipping±15% brightness change±10° rotation<b>Formats Provided:</b>COCO formatPyTorch YOLOV8 format<b>Image Size:</b> 416x416 pixels<b>Total Images:</b> 16,834Training: 13,862Validation: 1,975Testing: 997<b>Instance Numbers:</b>Containers: Over 10,000Hands: Over 8,000

本数据集包含饮水活动场景下第一人称视角（egocentric view）的容器与手部目标检测带标注图像，同时提供YOLOV8格式与COCO格式数据集。<br>更多细节请参阅本团队相关论文。<br><b>用途：</b>用于目标检测模型的训练与测试。<br><b>数据内容：</b>来自1-20号受试者第1阶段的视频。<br><b>图像来源：</b>从1-20号受试者第1阶段的视频中提取得到。<br><b>补充图像：</b>约500张来自Roboflow开源数据集的手部/容器图像；约1500张背景空样本图像，来源包括VOC数据集与MIT室内场景识别数据集：其中1000张为'MIT室内场景识别'数据集的室内场景图像，剩余400张为VOC数据集的无关物体图像。<br><b>数据增强：</b>水平翻转、±15%亮度调整、±10°角度旋转。<br><b>提供格式：</b>COCO格式、PyTorch YOLOV8格式。<br><b>图像尺寸：</b>416×416像素。<br><b>总图像数量：</b>共计16834张，其中训练集13862张、验证集1975张、测试集997张。<br><b>实例数量：</b>容器类实例超10000个，手部类实例超8000个。