Temporal Needle descriptor
Overview
The ability to detect similar actions across videos can be very useful for real-world applications in many fields. However, this task is still challenging for existing systems, since videos that present the same action, can be taken from significantly different viewing directions, preformed by different actors and backgrounds and under various video qualities. Video descriptors play a significant role in these systems.
In this work we propose the "temporal-needle" descriptor which captures the dynamic behavior, while being invariant to viewpoint and appearance. The descriptor is computed using multi temporal scales of the video and by computing self-similarity for every patch through time in every temporal scale. The descriptor is computed for every pixel in the video. However, to find similar actions across videos, we consider only a small subset of the descriptors - the statistical significant descriptors. This allow us to find good correspondences across videos more efficiently. Using the descriptor, we were able to detect the same behavior across videos in a variety of scenarios.
We demonstrate the use of the descriptor in tasks such as temporal and spatial alignment, action detection and even show its potential in unsupervised video clustering into categories.
In this work we handled only videos taken with stationary cameras, but the descriptor can be extended to handle moving camera as well. Example of similar descriptors of corresponding points across two videos: Figure: The temporal-needle of corresponding space-time points in two videos that present the same action. (a) presents 7 frames from a tennis serve by Roger Federer, we picked two points, p1(t) and q1(t), in the center frame t. (b) display 7 frames from a tennis serve by some other player and the points p2(t) and q2(t) are the corresponding points of p1(t) and q1(t), respectively. (c) display the descriptors of these 4 points. The descriptors were computed with patch size 3x3 (smaller than the rectangles presented in (a) and (b)), temporal radius of R=3 and 3 temporal scales. Although the videos were taken from very different view points and with different players and backgrounds, the descriptors of corresponding points are similar. Download the code: (will be added soon)