Team RUC_AIM3 Technical Report at Activitynet 2020 Task 2: Exploring Sequential Events Detection for Dense Video Captioning
This work addresses the problem of detecting and captioning events in untrimmed videos for video analysis applications, representing an incremental improvement over existing methods.
The paper tackled dense video captioning by proposing a novel model for event sequence generation that directly generates event boundaries using bi-directional temporal dependencies, achieving a state-of-the-art METEOR score of 9.894 on the ActivityNet 2020 dataset.
Detecting meaningful events in an untrimmed video is essential for dense video captioning. In this work, we propose a novel and simple model for event sequence generation and explore temporal relationships of the event sequence in the video. The proposed model omits inefficient two-stage proposal generation and directly generates event boundaries conditioned on bi-directional temporal dependency in one pass. Experimental results show that the proposed event sequence generation model can generate more accurate and diverse events within a small number of proposals. For the event captioning, we follow our previous work to employ the intra-event captioning models into our pipeline system. The overall system achieves state-of-the-art performance on the dense-captioning events in video task with 9.894 METEOR score on the challenge testing set.