VIDEO QUESTION ANSWERING

Summary from the TLDR AI newsletter

Visual question answering has seen massive adoption over the past several months, and rapid progress. Largely enabled by pre-trained models such as Blip and clip. Applying these models to video is challenging due to the dramatically increased computational requirements. This work introduces a cascade of different models that dramatically improves the robustness of these systems by operating at different timescales. They use Blip2 and find strong performance.