Guided Attention for Next Active Object @ Ego4D Short Term Object Interaction Anticipation Challenge

In this technical report, we describe the Guided-Attention mechanism [1] based solution for the short-term anticipation(STA) challenge for the EGO4D challenge. It combines theobject detections, and the spatiotemporal features extractedfrom video clips, enhancing the motion and contextual information, and further decoding the object-centric and motion-centric information to address the problem of STA in egocen-tric videos. For the challenge, we build our model on top ofStillFast [2] with Guided Attention applied on fast network.Our model obtains better performance on the validation setand also achieves state-of-the-art (SOTA) results on the challenge test set for EGO4D Short-Term Object Interaction Anticipation Challenge.