Adversarial Training with OCR Modality Perturbation for Scene-Text Visual Question Answering

Open in new window