Prime Video’s work on sports field registration, recap/intro detection
Two Prime Video papers at the Winter Conference on Applications of Computer Vision (WACV) 2021 proposed neural models for enhancing video-streaming experiences.
In January 2021, Prime Video teams presented two papers at the IEEE Winter Conference on Applications of Computer Vision (WACV).
The first paper was on automatic sports field registration, which is about understanding the spatial relationship between the sports field and the broadcast video-camera. This technology is critical for a variety of downstream sports-watching experiences, including player spotlight, player-speed, distance-covered, yards-to-go in NFL games, and multi-player events visualization (for example, offside fouls in soccer). Prime Video’s technology has been evaluated on the top-5 US tier-1 sports and offers state-of-the-art accuracy.
The second paper was about automatically detecting introductions and recaps in movies and TV shows so that Prime Video customers can enjoy a more laid-back video watching experience. To this end, Prime Video teams proposed a neural network that can distinguish the signature audio-visual patterns often found in intros and recaps. Unlike alternative approaches that require an entire video series to find intro and recap timestamps, our approach can work on each episode independently, which makes it more practical.
For more in-depth information about these topics, see the Prime Video’s work on sports field registration, recap/intro detection article on the Amazon Science website.