ICME2025 Audio Encoder Challenge

The ICME2025 Audio Encoder Challenge, hosted by Xiaomi, University of Surrey, and Dataocean AI, invites participants to evaluate audio encoders in real-world tasks. Register by April 1, 2025.

The IEEE International Conference on Multimedia & Expo (ICME) 2025 is hosting the Audio Encoder Capability Challenge, organized by Xiaomi, University of Surrey, and Dataocean AI. This challenge aims to rigorously evaluate audio encoders in real-world tasks.

Key features of the challenge include:

• No restrictions on model size or training data scale
• Allowed use of pre-trained models
• Comprehensive testing across speech, environmental sounds, and music
• Open-source evaluation system available at GitHub

Registration:

To participate, register by April 1, 2025, using the registration form. The challenge begins on February 7, 2025.

Submission:

• Clone the audio encoder template from GitHub
• Implement your audio encoder following the README.md instructions
• Submit a ZIP file of the repository and a technical report (PDF, max 6 pages) by April 30, 2025, to 2025icme-aecc at dataoceanai.com

For more details, visit the official challenge page.

Tags: ICME2025, Audio Encoder Challenge, IEEE, Xiaomi, University of Surrey, Dataocean AI, Machine Learning, Audio Processing