File size: 1,093 Bytes
84fd22f f6fa0db e524f1a f6fa0db |
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 |
---
language: ko
license: mit
tags:
- audio
- emotion-detection
- classification
metrics:
- accuracy
model-index:
- name: audio-emotion-model
results:
- task:
type: audio-classification
name: Audio Classification
dataset:
name: custom-dataset
type: custom
metrics:
- type: accuracy
value: 0.92
---
- Input: MFCC 13ch, length 100 → shape (B, 13, 100)
- Delta: (X - mean) / (std + 1e-8)
- Labels: see `labels.json` (index ↔ label 1:1)
## Usage
```python
import json, torch, numpy as np
from huggingface_hub import hf_hub_download
from importlib.machinery import SourceFileLoader
repo = "HyukII/audio-emotion-model"
w = hf_hub_download(repo, "pytorch_model.pth")
m = hf_hub_download(repo, "model.py")
lab = hf_hub_download(repo, "labels.json")
labels = json.load(open(lab, encoding="utf-8"))
Model = SourceFileLoader("amodel", m).load_module().PyTorchAudioModel
model = Model(num_labels=len(labels)).eval()
state = torch.load(w, map_location="cpu")
model.load_state_dict(state)
# x: tensor (1,13,100) → probs = softmax(model(x), dim=1) |