Ep. 24: How Yahoo Uses AI to Create Instant eSports Highlight Reels
May 31, 2017
auto_awesome
Yale Song, a Senior Research Scientist at Yahoo! Research, dives into the fascinating world of AI in eSports highlight reel generation. He discusses how AI technology is revolutionizing the immediate curation of electrifying gameplay moments, meeting the voracious demands of fans. Yale explains the challenges of highlighting key moments amid chaos, how deep learning classifies highlights, and how AI can even create shareable GIFs in real-time. The conversation highlights the shift towards objective analyses in sports, reshaping how we experience eSports.
AI is revolutionizing eSports by rapidly creating highlight reels, addressing the fans' demand for immediate and engaging content.
The implementation of objective visual indicators helps the AI to effectively identify significant gameplay moments, enhancing overall viewer satisfaction.
Deep dives
AI in eSports Highlight Reels
AI is being utilized to rapidly produce highlight reels for eSports, addressing the demand for immediate content from games like Starcraft and League of Legends. Yahoo's eSports team recognized the challenge of editing and broadcasting significant moments from numerous live-streamed games, which led them to consider AI as a solution. This approach allows the identification of key moments without hiring multiple editors, thereby significantly speeding up the content production process. By leveraging AI, broadcast companies can deliver exciting highlights in real time, enhancing viewer engagement.
Objective Criteria for Highlights
To tackle the subjectivity of highlight moments, the team implemented a system that identifies objective visual indicators within eSports games, such as dramatic visual effects like bursts of fire or explicit messages like 'red team kills'. This method helps reduce the ambiguity of what constitutes a 'highlight' by focusing on the common visual cues that signify significant gameplay actions. The team gathered over 100 hours of pro league gameplay footage for training the AI, ensuring it learns to recognize these objective components across different games. By establishing these criteria, the AI not only identifies potential highlights but does so in a way that maximizes viewer satisfaction.
High Performance Metrics of AI
The AI system demonstrated impressive performance metrics, achieving an average precision rate close to 90% and a recall rate near 80% when identifying highlights. This indicates a strong ability to accurately detect significant moments while minimizing false positives. The approach included a thorough testing phase using annotated videos to train the system, ensuring that it learns effectively from prior examples. The effectiveness of this technology allows for quick production of content, resulting in a more efficient workflow compared to traditional manual editing methods.
Whatever sport we follow, we all love a good highlight reel - and we want those highlights now. And whether they're following StarCraft II, League of Legends, or Heroes of the Storm, eSports fans are no different. To highlight the kills, and thrills, of a great eSports competition, Yale Song, Senior Research Scientist at Yahoo! Research, turned to AI.
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode