Scribd
Upload
5 views2 pages

Sound Classification with ESC-10 on Huawei Cloud

This guide provides instructions for downloading, organizing, and training a sound classification model using the ESC-10 dataset on Huawei Cloud ModelArts. It includes steps for preparing the dataset, training the model through the ExML interface, and optionally deploying the model for testing. The ESC-10 dataset is a subset of the ESC-50 dataset, suitable for quick experimentation in sound classification tasks.

Uploaded by

hccdauser162
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
5 views2 pages

Sound Classification with ESC-10 on Huawei Cloud

This guide provides instructions for downloading, organizing, and training a sound classification model using the ESC-10 dataset on Huawei Cloud ModelArts. It includes steps for preparing the dataset, training the model through the ExML interface, and optionally deploying the model for testing. The ESC-10 dataset is a subset of the ESC-50 dataset, suitable for quick experimentation in sound classification tasks.

Uploaded by

hccdauser162
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Guide: Sound Classification on Huawei Cloud ModelArts

using ESC-10 Dataset


This guide helps you download, organize, and train a sound classification model using the
ESC-10 dataset on Huawei Cloud ModelArts via the ExML (Experiential Machine Learning)
interface. ESC-10 is a small, well-known environmental sound dataset, ideal for quick
training and experimentation.

2. Download and Prepare ESC-10 Dataset in Notebook


Run the following code on notebook:

# Step 1: Download the ESC-50 dataset (ESC-10 is a subset)


wget [Link]

# Step 2: Unzip the dataset


unzip [Link]

Now run this code to extract only the ESC-10 subset and organize it into labeled folders:

import pandas as pd
import os
import shutil

# Load metadata
meta = pd.read_csv('ESC-50-master/meta/[Link]')

# Filter ESC-10 entries


esc10 = meta[meta['esc10'] == True]

# Create output folder


[Link]('esc10_dataset', exist_ok=True)

# Organize into labeled folders


for _, row in [Link]():
label = row['category']
src_path = f"ESC-50-master/audio/{row['filename']}"
dst_folder = f"esc10_dataset/{label}"
[Link](dst_folder, exist_ok=True)
[Link](src_path, dst_folder)
3. Zip the Dataset
Use the following command to zip the dataset for download or upload to OBS:

Moxing as mox

6. Train Sound Classification Model in ExML


Now go to ModelArts > Experiential ML (ExML) and follow these steps:

1. Click on “Create Project” → Select “Sound Classification”.


2. Choose “Custom Dataset” and select the OBS path: obs://your-bucket-
name/esc10_dataset/
3. Set training/validation split (e.g., 80/20).
4. Click “Next” → Choose model type and training configuration.
5. Start training and monitor accuracy/loss.

7. Deploy the Model (Optional)


After training, deploy the model with one click and get a REST API endpoint to test real
audio files.

You might also like