dharshanzeb/yolo-detection-dataset

这是一个用于YOLO目标检测模型训练的合成数据集，包含5个类别（汽车、人、狗、猫、自行车）。数据集分为训练集（500张图像）、验证集（100张图像）和测试集（50张图像），所有图像尺寸均为640×640。数据集中包含15%的硬负样本（无对象的图像），用于减少训练中的假阳性。数据集提供了两种格式：HuggingFace数据集格式（Parquet）和YOLO格式（Zips），并支持YOLO内置的马赛克和混合增强技术。A synthetic object detection dataset with 5 classes (car, person, dog, cat, bicycle), ready for training YOLOv8/v11 models. The dataset is split into train (500 images), validation (100 images), and test (50 images) sets, all with image sizes of 640×640. It includes 15% hard negatives (images with no objects) to reduce false positives during training. The dataset is available in two formats: HF Datasets (Parquet) and YOLO Format (Zips), and is designed for use with YOLOs built-in Mosaic and Mixup augmentations.