VQAv2 (Validation Set)

This repository contains the VQAv2 validation dataset organized into a structured directory layout for model evaluation on a course project.

• Project GitHub Repository: VLM-TTA
• Official Dataset Page: VisualQA v2 Official Site

---

Dataset Summary

• Split: Validation set (Balanced Real Images)
• Images: 40,504
• Questions: 214,354
• Annotations: 2,143,540

---

Directory & File Structure

The archive contains the following layout:

VQAv2/
├── images/
│   └── val2014/
│       ├── COCO_val2014_000000000042.jpg
│       └── ...
├── questions/
│   └── v2_OpenEnded_mscoco_val2014_questions.json
└── annotations/
    └── v2_mscoco_val2014_annotations.json

• Questions & Annotations Schema: Follows the original official VQAv2 JSON schema.

---

Results

Model predictions associated with this split in 4 variations of question(Q) subsets: 100Q, 1000Q, 10000Q, Full, can be found in the /results directory.

---

Citation & Attribution

Please cite the original publication when using this dataset:

@inproceedings{vqav2,
  title={Making the v in vqa matter: Elevating the role of image understanding in visual question answering},
  author={Goyal, Yash and Khot, Tejas and Summers-Stay, Douglas and Batra, Dhruv and Parikh, Devi},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2017}
}

• Official Paper: Making the v in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering

License

This dataset is distributed under the original authors' Creative Commons Attribution 4.0 International (CC-BY-4.0) license.