torinriley/FTCVision

FTC Vision: 2024-2025数据集是专为2024-2025赛季的FIRST Tech Challenge（FTC）目标检测任务设计的官方数据集。该数据集包含图像和注释，注释采用XML VOC格式，图像分为训练集和验证集。数据集特别优化了TensorFlow的使用，提供了TFRecord文件和标签映射文件，同时也提供了RAW格式以便在其他框架中使用。数据集的结构便于图像分类和模型评估，并且包含了不同颜色游戏件的精确边界框注释。

The FTC Vision: 2024-2025 Dataset for Object Detection is the official dataset designed for the FTC (FIRST Tech Challenge) 2024-2025 season. This dataset is specifically designed for training object detection models using TensorFlow, and it is also provided in RAW format for use in other frameworks or custom applications. The dataset contains images and annotations in XML VOC format, focusing on different-colored game pieces. The annotations are divided into train and val subsets, and TFRecord files and a label_map.pbtxt file are provided for TensorFlow users. The images are organized by class, ensuring compatibility with TensorFlows image_dataset_from_directory(). The dataset also supports other frameworks like PyTorch and is expandable, allowing for the addition of new game pieces or classes.