Who leads the DESED benchmark?

ATST-SED currently leads DESED with a score of 58.1 on Event F1.

What is the state-of-the-art score on DESED?

The state-of-the-art result on DESED is 58.1 (Event F1), achieved by ATST-SED as of 2026.

How many models are tracked on DESED?

Codesota tracks 3 models on DESED.

When was the DESED leaderboard last updated?

The DESED leaderboard on Codesota includes results through 2026.

Codesota · Benchmark · DESEDHome/Leaderboards/Audio & Speech/Sound Event Detection/DESED

Unknown

DESED.

Name: DESED Benchmark Results
Creator: Unknown
Published: 2026-01-01
License: https://creativecommons.org/licenses/by/4.0/

Sound event detection dataset covering 10 domestic sound classes. Used as the benchmark for DCASE Challenge Task 4. Primary metric is event-based macro-F1 on the public eval / validation set.

Paper ↗Leaderboard ↓

§ 01 · Leaderboard

Results by metric.

Only 3 models on this benchmark

Help build the community leaderboard — submit your model results.

Found a wrong score or missing run?

Use row edits to send a sourced correction into moderation.

Add / edit result ↗Report issue ↗

Event F1

Event F1 is the reported evaluation metric for DESED. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for Event F1verifiedpapervendorcommunityunverified

Rank	Model	Trust	Score	Year	Links	Fix
01	ATST-SED ATST-SED, DESED public eval, event-based F1. seed — verify.	paper	58.1	2026	Source ↗	Looks wrong?
02	FDY-CRNN FDY-CRNN, DESED validation, event-based F1. Nam et al. ICASSP 2022.	paper	45.1	2026	Source ↗	Looks wrong?
03	Prototypical CRNN (DCASE2021 baseline) DCASE 2021 Task 4 baseline, event-based F1. seed — verify.	paper	40.1	2026	Source ↗	Looks wrong?

§ 04 · Submit a result

Add to the leaderboard.

← Back to Sound Event Detection