Spaces:
Running on Zero
Running on Zero
File size: 2,845 Bytes
87094a8 d0ee797 87094a8 6881841 87094a8 2d5f79e 87094a8 2d5f79e 87094a8 2d5f79e 87094a8 2d5f79e 87094a8 2d5f79e 87094a8 2d5f79e 87094a8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 | ---
title: Audio Mood Classifier
emoji: 🎵
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: "5.33.0"
app_file: app.py
pinned: false
fullWidth: false
license: apache-2.0
short_description: Classify the mood of an uploaded song
suggested_hardware: zero-a10g
startup_duration_timeout: 45m
preload_from_hub:
- guyPerry/audio-mood-classifier
models:
- guyPerry/audio-mood-classifier
---
# 🎵 Audio Mood Classifier 🎵
> Can a machine feel the mood of a song?
Upload any song (MP3) and get an instant mood prediction from a fine-tuned deep learning model — no lyrics, no metadata, just the raw audio.
The model is a fine-tuned [Audio Spectrogram Transformer (AST)](https://huggingface.co/MIT/ast-finetuned-audioset-10-10-0.4593) pre-trained by MIT on AudioSet. Because mood is inherently subjective, it targets broad emotional character rather than precise genre.
**Links:** [Model card](https://huggingface.co/guyPerry/audio-mood-classifier) · [Project on GitHub](https://github.com/PerryGu/audio_mood_classifier_hf)
---
## How this demo works
For each upload, the app:
1. Skips the first and last **30 seconds** of the track (to avoid intros and outros).
2. Extracts **3 evenly-spaced clips**, **10 seconds** each, from the remaining audio.
3. Resamples to **16 kHz** mono and applies **EBU R128 loudness normalization** (−20 LUFS), matching the training pipeline.
4. Classifies each clip independently, then **combines the confidence scores** (late fusion) into one final prediction.
Training used **6 segments per song**; the demo uses **3** for a faster response while keeping the same sampling logic.
---
## Training dataset
The model was trained on a **custom dataset built from scratch** — a hand-curated catalog matched against a personal music library. Each song was manually assigned to one mood category.
| | |
|---|---|
| **Total songs** | 250 |
| **Mood categories** | 3 |
| **Segments per song (training)** | 6 |
| **Segment length** | 10 seconds |
| **Guard buffer** | 30 s skipped at each end of every track |
| **Total training segments** | up to 1,500 (250 songs × 6 segments) |
| **Audio preprocessing** | 16 kHz mono · EBU R128 loudness norm (−20 LUFS) |
| **Train / eval / test split** | ~70% / 15% / 15% (group-aware — all segments from the same song stay in the same split) |
### Songs per category
| Category | Songs in catalog |
|---|---|
| 🌧 `calm_melancholic` | 84 |
| 😐 `moderate_neutral` | 74 |
| ⚡ `energetic_upbeat` | 90 |
| **Total** | **250** |
---
## Mood classes
| Label | Description |
|---|---|
| 🌧 `calm_melancholic` | Slow, introspective, melancholic |
| 😐 `moderate_neutral` | Balanced, mid-energy, neutral feel |
| ⚡ `energetic_upbeat` | Fast, high-energy, upbeat |
|