Open-source model of baby-cry-classification-finetuned-babycry-v4

Home

Baby Cry Classification Finetuned Babycry V4

Developed by Wiam

A baby cry classification model fine-tuned based on wav2vec2-large-xlsr-53-english, achieving 81.5% accuracy

Audio Classification

Transformers

Open Source License:Apache-2.0 #Baby cry recognition #Audio classification #Wav2Vec2 fine-tuning

Downloads 120

Release Time : 10/1/2024

Model Overview

This is an audio classification model specifically designed to recognize and classify baby cries. It is fine-tuned from the jonatasgrosman/wav2vec2-large-xlsr-53-english model and can be used in baby monitoring and health monitoring scenarios.

Model Features

High accuracy classification

Achieves 81.5% accuracy on the evaluation set, effectively recognizing baby cries

Based on wav2vec2 architecture

Utilizes the powerful wav2vec2-large-xlsr-53 as the base model, with excellent audio feature extraction capabilities

Lightweight fine-tuning

Requires only a small amount of training data to achieve good performance, suitable for specific scenario applications

Model Capabilities

Baby cry recognition

Audio classification

Sound event detection

Use Cases

Baby monitoring

Smart baby monitoring system

Integrated into baby monitoring devices to detect baby cries in real-time

Helps parents or caregivers respond promptly to baby's needs

Baby health monitoring

Analyzes cry characteristics to assess possible health conditions

Assists in determining if the baby is uncomfortable or requires medical attention

Medical assistance

Newborn ward monitoring

Automatically monitors the frequency and patterns of baby cries in hospital newborn wards

Helps medical staff assess baby's condition

🚀 baby-cry-classification-finetuned-babycry-v4

This model is a fine - tuned solution for baby cry classification. It is based on the pre - trained model jonatasgrosman/wav2vec2-large-xlsr-53-english and trained on the audiofolder dataset. It offers high - accuracy classification of baby cries, providing valuable insights for related applications.

🚀 Quick Start

This model is a fine - tuned version of jonatasgrosman/wav2vec2-large-xlsr-53-english on the audiofolder dataset. It achieves the following results on the evaluation set:

Loss: 0.7255
Accuracy: {'accuracy': 0.8152173913043478}
F1: 0.7322
Precision: 0.6646
Recall: 0.8152

📚 Documentation

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 4
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.6244	0.5435	25	0.7271	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152
0.6901	1.0870	50	0.7196	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152
0.5873	1.6304	75	0.7426	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152
0.8029	2.1739	100	0.7124	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152
0.5661	2.7174	125	0.7259	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152
0.6121	3.2609	150	0.7431	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152
0.7571	3.8043	175	0.7316	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152
0.5284	4.3478	200	0.7277	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152
0.7182	4.8913	225	0.7255	{'accuracy': 0.8152173913043478}	0.7322	0.6646	0.8152

Framework versions

Transformers 4.44.2
Pytorch 2.4.1+cu121
Datasets 3.0.1
Tokenizers 0.19.1

📄 License

This project is licensed under the Apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご