Optimizing Through Learned Errors for Accurate Sports Field Registration
This work addresses accurate sports field registration for broadcast video analysis, presenting an incremental improvement over existing methods.
The paper tackles the problem of registering sports field templates onto broadcast videos by training a deep network to regress registration errors and optimizing parameters to minimize these errors, outperforming state-of-the-art methods on real-world videos and showing gains in synthetic scenarios.
We propose an optimization-based framework to register sports field templates onto broadcast videos. For accurate registration we go beyond the prevalent feed-forward paradigm. Instead, we propose to train a deep network that regresses the registration error, and then register images by finding the registration parameters that minimize the regressed error. We demonstrate the effectiveness of our method by applying it to real-world sports broadcast videos, outperforming the state of the art. We further apply our method on a synthetic toy example and demonstrate that our method brings significant gains even when the problem is simplified and unlimited training data is available.