Dataset Card for synthetic us passports This is a FiftyOne dataset with 9750 samples. Installation If you haven't already, install FiftyOne: Usage Based on the information you've provided, here's a filled out dataset card for your FiftyOne dataset: Dataset Details Dataset Description This is a synthetic dataset of US passport images designed to test Vision Language Models (VLMs) on document understanding tasks. The dataset challenges models with realistic scenarios including tilted documents and high resolution images where the passport occupies only a small region of interest. Each sample contains a synthetically generated US passport with complete biographical and document information fields. The dataset is particularly useful for evaluating model robustness in document AI applications where documents may not be perfectly aligned or centered in the frame. Curated by: Arnaud Stiegler Language(s) (NLP): en License: Apache 2.0 Dataset Sources Repository: https://github.com/arnaudstiegler/synth doc AI Original Dataset: https://huggingface.co/datasets/arnaudstiegler/synthetic us passports easy Uses Direct Use This dataset is intended for: Training and evaluating Vision Language Models…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy