Chouoftears/Single-DriveLM-NuScenes-VQA

Single DriveLM NuScenes VQA Dataset

[2024.11.10]VQA test set is now released!!!

Labels are provided

Dataset Description

This is the sub-dataset of DriveLM which only include single object in ego scenes

Uses

For single traffic participant recgonition, segmentation, VQA subtasks of driving scenarios.

Dataset Structure

single_pedestrian 
├── images  
└── labeled_pedestrian_data.json

single_vehicle
├── images  
└── labeled_vehicle_data.json

Labels Structure

Pedestrian  
├── token   
├── image_path   
└── labeled_information   
    ├── gender   
    ├── age_group  
    └── race  

Vehicle  
├── token   
├── image_path   
└── labeled_information   
    ├── type 
    └── color  

Original Dataset Sources

• Repository: https://github.com/OpenDriveLab/DriveLM
• Paper: https://arxiv.org/abs/2312.14150

Dataset Card Authors

Shenzhe Zhu