File size: 2,639 Bytes
fda6e40 e2dd32d fda6e40 4c6f1dd fda6e40 e2dd32d fda6e40 e2dd32d fda6e40 e2dd32d c6bd35b e2dd32d 5b7288e e2dd32d c6bd35b fda6e40 04ac83f fda6e40 b3653e3 fda6e40 |
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 |
import gradio as gr
from transformers import pipeline
import numpy as np
import os
from huggingface_hub import login
import librosa
import spaces
HF_TOKEN = os.environ.get("HF_TOKEN")
if HF_TOKEN:
login(token=HF_TOKEN)
MODEL_ID = "badrex/w2v-bert-2.0-kinyarwanda-asr"
transcriber = pipeline("automatic-speech-recognition", model=MODEL_ID)
@spaces.GPU
def transcribe(audio):
sr, y = audio
# convert to mono if stereo
if y.ndim > 1:
y = y.mean(axis=1)
# resample to 16kHz if needed
#if sr != 16000:
# y = librosa.resample(y, orig_sr=sr, target_sr=16000)
y = y.astype(np.float32)
y /= np.max(np.abs(y))
return transcriber({"sampling_rate": sr, "raw": y})["text"]
examples = []
examples_dir = "examples"
if os.path.exists(examples_dir):
for filename in os.listdir(examples_dir):
if filename.endswith((".wav", ".mp3", ".ogg")):
examples.append([os.path.join(examples_dir, filename)])
print(f"Found {len(examples)} example files")
else:
print("Examples directory not found")
demo = gr.Interface(
fn=transcribe,
inputs=gr.Audio(),
outputs="text",
title="<div>ASRwanda 🎙️ <br>Speech Recognition for Kinyarwanda</div>",
description="""
<div class="centered-content">
<div>
<p>
Developed with ❤ by <a href="https://badrex.github.io/" style="color: #2563eb;">Badr al-Absi</a> ☕
</p>
<br>
<p style="font-size: 15px; line-height: 1.8;">
Muraho 👋🏼
<br>
<br>
This is a demo for ASRwanda, a Transformer-based automatic speech recognition (ASR) system for Kinyarwanda language.
The underlying ASR model was trained on 500 hours of transcribed speech provided by
<a href="https://digitalumuganda.com/" style="color: #2563eb;">Digital Umuganda</a> as part of the Kinyarwanda
<a href="https://www.kaggle.com/competitions/kinyarwanda-automatic-speech-recognition-track-a" style="color: #2563eb;"> ASR hackathon</a> on Kaggle.
<br>
<p style="font-size: 15px; line-height: 1.8;">
Simply <strong>upload an audio file</strong> 📤 or <strong>record yourself speaking</strong> 🎙️⏺️ to try out the model!
</p>
</div>
</div>
""",
examples=examples if examples else None,
cache_examples=False,
flagging_mode=None,
)
if __name__ == "__main__":
demo.launch() |