6 Jan, 16:00, online
Ali Mohammad, PhD student
Different perspective on text/image matching
Cross-modality information retrieval is a popular research task that grew more important as cross-modality data becoming more common on the internet, In my seminar talk, I will discuss the usage of transformer architecture for textual and visual modalities and training tasks that can be used to improve its performance, including generative tasks for both textual and visual modalities. Finally, I will discuss a new approach for collecting data from videos for text/image retrieval tasks.