UniParser/MolDet-Bench

分子检测基准集合是一个专注于化学领域目标检测任务的数据集。它包括两个主要任务：多尺度化学结构检测（MolDet-Bench-General）和文档中的化学结构检测（MolDet-Bench-Doc）。该数据集旨在评估分子定位算法在各种现实条件下的性能，包括不同尺度和文档类型。数据集还整合了第三方基准BioVista以进行额外评估。基准以Ultralytics YOLO格式组织，并包括mAP50和mAP50-95等评估指标。README提供了每个基准的详细描述，包括图像和分子数量、标注方法以及评估结果。

The Molecule Detection Benchmark Collection is a dataset focused on object detection tasks in the chemistry domain. It includes two main tasks: multi-scale chemical structure detection (MolDet-Bench-General) and chemical structure detection in documents (MolDet-Bench-Doc). The dataset is designed to evaluate the performance of molecular localization algorithms under various real-world conditions, including different scales and document types. The dataset also incorporates a third-party benchmark, BioVista, for additional evaluation. The benchmarks are organized in the Ultralytics YOLO format and include evaluation metrics such as mAP50 and mAP50-95. The README provides detailed descriptions of each benchmark, including the number of images and molecules, annotation methods, and evaluation results.