Sidewalk_road_yolo_dataset.7z

<b>Dataset (Image) Collection:</b>A dataset of sidewalk and road images is downloaded from roboflow, which is an end-to-end computer vision platform that simplifies the process of building computer vision models.The dataset is a semantic segmentation dataset out of which only the images have been used to create a new dataset because the requirement of the project is to obtain sidewalk and road areas in terms of small bounding boxes, rather than overall bounding box.<br><b>b.</b><b> </b><b>Dataset Annotation</b>As for reasons stated in previous section, the images obtained have been annotated manually one by one using a tool, labelImg, which is a sophisticated open-source tool crafted for the purpose of annotating objects present in images. It streamlines the task of delineating bounding boxes around these objects and associating them with class labels. Individuals have the capability to sketch these bounding boxes, store annotations in file types such as PASCAL VOC, YOLO, or CreateML, and leverage them for a range of computer vision assignments. This Python-driven tool is adaptable across various platforms and is extensively employed by developers and researchers alike for the annotation of datasets and the training of models.

<b>图像数据集集合：</b>本数据集的人行道与道路图像均下载自Roboflow——一款可简化计算机视觉模型构建流程的端到端计算机视觉平台。该数据集原本为语义分割数据集，本次仅提取其中的图像以构建新数据集，原因在于本项目需求为获取人行道与道路区域的小型边界框（bounding box），而非整体边界框。<br><b>b.</b><b> </b><b>数据集标注</b>基于前文所述缘由，我们使用标注工具labelImg对获取到的图像逐张进行手动标注。labelImg是一款专为图像内目标标注打造的专业开源工具，可简化为目标绘制边界框并关联类别标签的流程。用户可手动绘制边界框，并将标注结果存储为PASCAL VOC、YOLO或CreateML等格式，以支持各类计算机视觉任务。这款基于Python开发的工具兼容多平台，被开发者与研究者广泛用于数据集标注与模型训练流程中。