Fetching the paper…

VoLTA: Vision-Language Transformer with Weakly-Supervised Local-Feature Alignment · Around