---
title: Overlap-Aware Sound Event Detection
emoji: 🔊
colorFrom: blue
colorTo: purple
sdk: gradio
app_file: app.py
pinned: false
sdk_version:  5.49.1
---

# Overlap-Aware Sound Event Detection

This Hugging Face Space ports the inference pipeline from `sed_ensemble_v2.ipynb`.

- AST × 3 and PaSST AudioSet-527 classifiers
- Family-level fusion and agreement gating
- Optional CLAP-based extra category prompts
- Optional Silero VAD to refine speech boundaries
- Timeline plot showing overlapping sound events

The DCASE 2024 CRNN checkpoint-based `app.py` is a separate system and is not used by this Space.

By default, audio is limited to 60 seconds for hosted CPU practicality. Set the Space variable `MAX_SEC` to a different integer (for example `180`) to change this limit.