Discover our 2027 challenges

The SoccerNet Challenges return for their 7th edition, with four tasks advancing computer vision and AI for football: Group Activity Recognition, Streaming Visual Question Answering, Long Form Game State Reconstruction, and Novel View Synthesis.

Each task will launch independently as its data, baseline and evaluation benchmark become ready. The submission deadline is April 25, 2027, 23:59 Anywhere on Earth (UTC−12).

Stay connected for task announcements, tutorials and updates. Join the discussion on Discord and explore our YouTube channel to get started.

Whether you work on action understanding, 3D reconstruction, player tracking or vision and language, join us in advancing AI for football!

CHAT WITH US ON DISCORD

Group Activity Recognition 2027

📢 1️⃣ Group Activity Recognition Challenge! 🚀

How do pixels and player positions contribute to understanding collective activities in football? Building on SN-GAR, this challenge explores group activity recognition using visual and positional information. Participants may use pixels, positions, or both. The reference baseline uses 16 sampled frames from the tracking branch; the videos and tracking-full branches are also available.

🗂️Data / 💻Development Kit / 📚Rules / 📊Eval

🤝Sponsor: TBD / 💰Prize: TBD

Streaming Visual Question Answering 2027 (Coming Soon...)

📢 2️⃣ Streaming Visual Question Answering Challenge!

Can AI answer questions about a match as it unfolds? This challenge brings together football video understanding and language, asking models to answer questions using information available in an incoming video stream. The focus is on understanding what has been observed so far, without relying on future footage.

🗂️Data / 💻Development Kit / 📚Rules / 📊Eval

🤝Sponsor: TBD / 💰Prize: TBD

Long Form Game State Reconstruction 2027 (Coming Soon...)

📢 3️⃣ Long Form Game State Reconstruction Challenge!

Understanding a match requires following the game over time. This challenge focuses on reconstructing player positions and identities on the pitch across extended football sequences. Participants will maintain a coherent game state as players move, interact, and become occluded across one continuously annotated match half.

🗂️Data / 💻Development Kit / 📚Rules / 📊Eval

🤝Sponsor: TBD / 💰Prize: TBD

Novel View Synthesis 2027 (Coming Soon...)

🔉4️⃣ Novel View Synthesis Challenge! 🚀

Can AI show the game from a new perspective? This challenge explores the generation of unseen viewpoints from football footage. Participants will reconstruct the scene and render new views of players and the pitch using a more realistic simulator and fewer cameras than in 2026.

🗂️Data / 💻Development Kit / 📚Rules / 📊Eval

🤝Sponsor: TBD / 💰Prize: TBD