Hi, thanks for the open-source of this great work! However, I find some of the annotations in the datasets are inaccurate. For example, it's easy to find the color of the traffic light is inconsistent with the image, which could impede the performance when training with wrong answers. It would be great if the annotations could be filtered and corrected again.
Thank you for your issue. It would be great if you could specify which dataset and which scenarios you are referring to. For DriveLM-nuScenes, the traffic light information comes from OpenLane-V2.
Hi, thanks for the open-source of this great work! However, I find some of the annotations in the datasets are inaccurate. For example, it's easy to find the color of the traffic light is inconsistent with the image, which could impede the performance when training with wrong answers. It would be great if the annotations could be filtered and corrected again.