A clean, deduplicated and leakage-free Visual-Question-Answering set for
Vietnamese traffic signs, split train/test by the original detection split.
This is a curated derivative — it is not the same as
Anakonkai/vietnamese-traffic-sign-vqa:
that source is dominated by near-identical video frames and has no clean
train/test boundary. Here those issues are fixed (see How it was built).
What's inside… See the full description on the dataset page: https://huggingface.co/datasets/tungitachi/vietnamese-traffic-sign-vqa-dedup-1k.