LMDrive

该数据集包含了大约64,000个遵循指令的数据片段，以及464,000个注意指令，旨在训练智能驾驶代理根据多模态输入生成行动。该数据集利用CARLA模拟器进行收集，涵盖了多样的环境条件，并包含了各种导航和注意指令。规模上，数据集包含了64,000个解析过的片段和464,000个注意指令。任务方面，数据集专注于闭环端到端的自动驾驶。

This dataset contains approximately 64,000 instruction-following data segments and 464,000 attention instructions, designed to train autonomous driving agents to generate driving actions based on multimodal inputs. Collected using the CARLA simulator, this dataset covers diverse environmental conditions and incorporates a wide range of navigation and attention instructions. Regarding dataset scale, it comprises 64,000 parsed data segments and 464,000 attention instructions. This dataset focuses on closed-loop end-to-end autonomous driving tasks.