Weakly Supervised Temporal Action Localization Using Deep Metric Learning
This addresses the high annotation cost for video action localization, making it more scalable for real-world applications, though it is an incremental improvement over existing weakly supervised methods.
The paper tackles the problem of temporal action localization in videos by proposing a weakly supervised method that only requires video-level action labels, eliminating the need for expensive temporal annotations. It achieves a 6.5% mAP improvement on THUMOS14 at IoU threshold 0.5 and competitive results on ActivityNet1.2.
Temporal action localization is an important step towards video understanding. Most current action localization methods depend on untrimmed videos with full temporal annotations of action instances. However, it is expensive and time-consuming to annotate both action labels and temporal boundaries of videos. To this end, we propose a weakly supervised temporal action localization method that only requires video-level action instances as supervision during training. We propose a classification module to generate action labels for each segment in the video, and a deep metric learning module to learn the similarity between different action instances. We jointly optimize a balanced binary cross-entropy loss and a metric loss using a standard backpropagation algorithm. Extensive experiments demonstrate the effectiveness of both of these components in temporal localization. We evaluate our algorithm on two challenging untrimmed video datasets: THUMOS14 and ActivityNet1.2. Our approach improves the current state-of-the-art result for THUMOS14 by 6.5% mAP at IoU threshold 0.5, and achieves competitive performance for ActivityNet1.2.