Published June 1, 2026 | Version v1
Dataset Open

Audio Moment Retrieval from Long Audio DCASE 2026 Evaluation dataset

Authors/Creators

  • 1. LY Corporation

Description

What is this?

This repository contains data for DCASE Challenge 2026 Task 6. 

  • Audio and text features using CLAP and a sliding window, following the same feature extraction protocol as the CASTELLA dataset.
  • submission template

File structure

clap 
   └──dcase2026_evaluation_audio_{vid}.npz
clap_text
   └──qiddcase2026_evaluation_q{qid}.npz

Raw audio files

If participants require the raw audio, please contact the organizers

Submission

  1. Predict the relevant moments for each qid in the submission template.
    • The audio and text features in this repository correspond to the qid and vid in the template.

  2. Fill in the predicted moments in the pred_relevant_windows field.
    • Multiple moments can be submitted, which will be used to compute the mAP.
    • Moments should be listed in descending order of confidence, and only the first moment in the list will be used for the final ranking.

Files

Files (65.5 MB)

Name Size Download all
md5:8337ca832788bf0046ce00043e487809
61.6 MB Download
md5:99ec441e7ab4912dc756a2eeb543d2f2
3.9 MB Download
md5:de3c76a8ced437b9ccd45c2e761cbab7
23.3 kB Download
md5:fe6e544283733786bd7fcfee0260648f
27.5 kB Download