Leaving Some Stones Unturned: Dynamic Feature Prioritization for Activity Detection in Streaming Video

– Qualitative examples showing recognition episodes in streaming environment. Also see our project webpage for video examples. (A) – Streaming recognition results with different detector speeds (B). – Un-trimmed detection results with different object detector speeds (C). – State feature and reward function design (D). – Details for policy iteration (E). – Details for observation imputation in batch environment (F). – Implementation details (G). – Details for un-trimmed video generation (H).