Detecting Attended Visual Targets in Video

https://github.com/ejcgt/attention-target-detection Figure 1: Visual attention target detection over time. We propose to solve the problem of identifying gaze targets in video. The goal of this problem is to predict the location of visually attended region (circle) in every frame, given a track of an individual's head (bounding box). It includes the cases where such target is out of frame (row-col: 1-2, 1-3, 2-1), in which case the model should correctly infer its absence.