三维场景重建路口相机图像分割数据

路侧基础设施不同路口相机画面图像分割数据集，将脱敏后图像中的行人、汽车、卡车、自行车及其他物体识别和分割。可用于三维模型重建场景以及相关算法训练数据，重建后的真实场景可用于自动驾驶算法的训练，用路侧视角识别的模型来提升自动驾驶算法的准确度。路口路侧视频通过人工筛选截取图像，定义图像的编号、图像的名称，获取图像的宽度和长度，通过模型识别出图像中的人、车、路、交通标识等元素，并经过人工优化，得到上述交通元素的编号、类别编号，获取到元素的标注框、标注的区域的面积，是否标注的是单个对象以及对应的像素分割点。得到这些数据后，可以确定某个元素在该图像中的位置以及图像的类别，可以用于三维重建算法的训练集，重建后的真实场景可用于自动驾驶算法的训练，用路侧视角识别的模型来提升自动驾驶算法的准确度。

Image Segmentation Dataset for Camera Footage Collected at Intersections of Roadside Infrastructure. This dataset identifies and segments pedestrians, cars, trucks, bicycles and other objects within de-identified images. It can be used for 3D model reconstruction scenarios and as training data for relevant algorithms. The reconstructed real-world scenes can be applied to train autonomous driving algorithms, and models trained from the roadside perspective can effectively improve the accuracy of autonomous driving algorithms. Images are extracted from roadside videos at intersections through manual screening, with image IDs, image names, width and height of each image recorded. Elements including pedestrians, vehicles, roads and traffic signs in the images are first identified by models, followed by manual optimization to obtain the IDs, category IDs, bounding boxes, area of annotated regions, whether the annotation corresponds to a single object, and the corresponding pixel segmentation points of the aforementioned traffic elements. With these data acquired, the position of a specific element in the image and the category of the image can be determined. This dataset can serve as a training set for 3D reconstruction algorithms, and the reconstructed real-world scenes can be used to train autonomous driving algorithms, with models trained from the roadside perspective further enhancing the accuracy of autonomous driving algorithms.