Fusion of Short-term and Long-term Attention for Video Mirror Detection
This work addresses the challenge of video mirror detection, which is incremental by extending static image methods to incorporate temporal consistency for improved accuracy.
The paper tackles the problem of detecting mirrors in videos by fusing short-term appearance features and long-term context information, achieving state-of-the-art performance on a new benchmark dataset of 19,255 frames from 281 videos.
Techniques for detecting mirrors from static images have witnessed rapid growth in recent years. However, these methods detect mirrors from single input images. Detecting mirrors from video requires further consideration of temporal consistency between frames. We observe that humans can recognize mirror candidates, from just one or two frames, based on their appearance (e.g. shape, color). However, to ensure that the candidate is indeed a mirror (not a picture or a window), we often need to observe more frames for a global view. This observation motivates us to detect mirrors by fusing appearance features extracted from a short-term attention module and context information extracted from a long-term attention module. To evaluate the performance, we build a challenging benchmark dataset of 19,255 frames from 281 videos. Experimental results demonstrate that our method achieves state-of-the-art performance on the benchmark dataset.