RoadTextVQA

Ground Truth Format

{
  "dataset_name": "RoadTextVQA",
  "dataset_version": 1.0,
  "data": [
    {
      "questionId": 6, // Unique identifier for each question
      "question": "What is the name of the shop that is seen before the Gregorys coffee, on the same side?", // The question string
      "answer": ["cava"], // List of possible answers
      "video": "19.mp4", // Filename of the video clip
      "videoId": 21, // Unique identifier for each video
      "split": "val" // Dataset split ["train", "val", "test"]
    },
    ...
  ]
}

Download Links

Annotations

• Train: train.json
• Val: val.json
• Test: test.json

wget http://cvit.iiit.ac.in/images/datasets/RoadTextVQA/train.json
wget http://cvit.iiit.ac.in/images/datasets/RoadTextVQA/val.json
wget http://cvit.iiit.ac.in/images/datasets/RoadTextVQA/test.json

Videos

• Videos: videos.zip

wget http://cvit.iiit.ac.in/images/datasets/RoadTextVQA/videos.zip

OCR

• OCR: ocr.zip

wget http://cvit.iiit.ac.in/images/datasets/RoadTextVQA/ocr.zip

Citation

@inproceedings{tom2023reading,
  title={Reading Between the Lanes: Text VideoQA on the Road},
  author={Tom, George and Mathew, Minesh and Garcia-Bordils, Sergi and Karatzas, Dimosthenis and Jawahar, CV},
  booktitle={International Conference on Document Analysis and Recognition},
  pages={137--154},
  year={2023},
  organization={Springer}
}