Notice

[2024-1-21] The blind testing phase is extended by one day, ending on Jan. 22 23:59:59 (AoE).

[2024-1-20] The blind test set was released here.

[2024-1-20] The clean speech and metadata of the non-blind test set were released.

[2024-1-4] The brief description of the blind test set is available in the Data page.

[2024-12-30] The brief description of the non-blind test set is available in the Data page.

[2024-12-30] The clean speech and metadata of the official validation set are released.

[2024-12-25] The leaderboard is swithed to the non-blind test phase and the non-blind test set has been released. It can be downloaded here.

[2024-12-23] We have updated the Rules page to clarify the policy regarding the use of pre-trained speech enhancement models. Using pre-trained speech enhancement models trained on data other than the official Challenge dataset is NOT allowed. Refer to Rules page for more details.

[2024-12-23] To accommodate the short challenge period, we have decided to extend the challegnge period for one week. The blind-test phase, originally scheduled for Jan. 13–15, is now rescheduled to Jan. 20–22. Due to the shortened evaluation timeline, the subjective evaluation protocol will change. Please refer to Timeline page for more details.

[2024-11-26] We have released the checkpoint of the baseline TF-GridNet model at HuggingFace.

[2024-11-26] We have opened the leaderboard and released the official valdation set for leaderbord submission. Refer to Submission and Data pages for more details.