CAR -- Cityscapes Attributes Recognition A Multi-category Attributes Dataset for Autonomous Vehicles

Self-driving vehicles are the future of transportation. With current\nadvancements in this field, the world is getting closer to safe roads with\nalmost zero probability of having accidents and eliminating human errors.\nHowever, there is still plenty of research and development necessary to reach a\nlevel of robustness. One important aspect is to understand a scene fully\nincluding all details. As some characteristics (attributes) of objects in a\nscene (drivers' behavior for instance) could be imperative for correct decision\nmaking. However, current algorithms suffer from low-quality datasets with such\nrich attributes. Therefore, in this paper, we present a new dataset for\nattributes recognition -- Cityscapes Attributes Recognition (CAR). The new\ndataset extends the well-known dataset Cityscapes by adding an additional yet\nimportant annotation layer of attributes of objects in each image. Currently,\nwe have annotated more than 32k instances of various categories (Vehicles,\nPedestrians, etc.). The dataset has a structured and tailored taxonomy where\neach category has its own set of possible attributes. The tailored taxonomy\nfocuses on attributes that is of most beneficent for developing better\nself-driving algorithms that depend on accurate computer vision and scene\ncomprehension. We have also created an API for the dataset to ease the usage of\nCAR. The API can be accessed through https://github.com/kareem-metwaly/CAR-API.\n