RACER-Mini
收藏资源简介:
RACER-Mini 是一个专为自动驾驶中视觉-语言-动作(VLA)模型训练设计的推理字幕数据集。该数据集包含约1,000个样本,每个样本由以下三部分组成:1) 前摄像头图像的时间序列;2) 自我车辆的未来轨迹;3) 对应的推理说明文本。数据集中所有图像均已通过Dashcam Anonymizer工具对人脸和车牌进行了匿名化处理以保护隐私。该数据集采用CC BY-NC-SA 4.0许可协议发布,由日本新能源产业技术综合开发机构(NEDO)资助的项目JPNP20017开发,并使用Qwen团队开发的Qwen3-VL系列模型进行标注。
RACER-Mini is a reasoning captioning dataset specifically designed for training vision-language-action (VLA) models in autonomous driving scenarios. This dataset contains approximately 1,000 samples, with each sample comprising three components: 1) a temporal sequence of front-facing camera images; 2) the future trajectory of the ego vehicle; 3) corresponding reasoning explanation texts. All images in the dataset have been anonymized for faces and license plates via the Dashcam Anonymizer tool to protect personal privacy. This dataset is released under the CC BY-NC-SA 4.0 license, developed under project JPNP20017 funded by the New Energy and Industrial Technology Development Organization (NEDO) of Japan, and annotated using the Qwen3-VL series models developed by the Qwen team.



