DistilHubert - Finetuned - GTZAN open - source audio classification model, fine - tuned with the GTZAN dataset, achieving a classification accuracy of 88%

Distilhubert Finetuned Gtzan

Developed by timjwhite

An audio classification model fine-tuned on the GTZAN music classification dataset based on the DistilHuBERT architecture, achieving 88% accuracy

Audio Classification

Transformers

Open Source License:Apache-2.0 #Music Classification #High Accuracy Audio Recognition #Lightweight HuBERT

Downloads 16

Release Time : 7/11/2023

Model Overview

This model is a variant of DistilHuBERT, specifically fine-tuned for music genre classification tasks, suitable for audio signal processing and analysis scenarios.

Model Features

Efficient Distilled Architecture

Lightweight distilled version based on HuBERT, reducing computational resource requirements while maintaining performance

High Accuracy

Achieves 88% classification accuracy on the GTZAN test set

Fast Inference

Faster inference speed compared to the original HuBERT model

Model Capabilities

Music Genre Classification

Audio Feature Extraction

Music Content Analysis

Use Cases

Music Information Retrieval

Automatic Classification for Music Streaming Platforms

Automatically adds genre labels to uploaded music

Automatic classification with 88% accuracy

Music Recommendation Systems

Content-Based Music Recommendation

Generates recommendations by analyzing audio features

🚀 distilhubert-finetuned-gtzan

This model is a fine - tuned version of Sandiago21/distilhubert-finetuned-gtzan on the GTZAN dataset. It offers a valuable solution for audio classification tasks, achieving high accuracy on the evaluation set.

📚 Documentation

Model Information

Property	Details
Model Type	Fine - tuned version of Sandiago21/distilhubert-finetuned-gtzan
Training Data	marsyas/gtzan
Metrics	Accuracy

This model achieves the following results on the evaluation set:

Loss: 0.9951
Accuracy: 0.88

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

🔧 Technical Details

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.0951	1.0	57	0.5566	0.87
0.0629	2.0	114	0.6819	0.83
0.0231	3.0	171	0.6118	0.86
0.0159	4.0	228	0.9208	0.83
0.0374	5.0	285	0.8746	0.85
0.1714	6.0	342	0.6671	0.87
0.2148	7.0	399	1.1850	0.79
0.0147	8.0	456	1.0551	0.79
0.0788	9.0	513	1.5179	0.79
0.0015	10.0	570	1.3290	0.8
0.0049	11.0	627	1.0943	0.85
0.0012	12.0	684	1.0667	0.85
0.0043	13.0	741	1.1816	0.82
0.0015	14.0	798	0.9108	0.88
0.0011	15.0	855	1.0289	0.87
0.001	16.0	912	0.7696	0.87
0.0006	17.0	969	0.8539	0.87
0.1001	18.0	1026	1.1917	0.78
0.0017	19.0	1083	1.0016	0.83
0.0525	20.0	1140	0.9513	0.88
0.0004	21.0	1197	0.9268	0.86
0.0003	22.0	1254	1.1209	0.82
0.0003	23.0	1311	0.9270	0.87
0.0003	24.0	1368	1.1148	0.84
0.0003	25.0	1425	1.0507	0.85
0.0002	26.0	1482	1.0156	0.86
0.0002	27.0	1539	1.0062	0.87
0.0002	28.0	1596	1.0124	0.87
0.0002	29.0	1653	1.0154	0.87
0.0002	30.0	1710	1.0092	0.88
0.0002	31.0	1767	1.0123	0.88
0.0175	32.0	1824	0.9928	0.88
0.0002	33.0	1881	1.0014	0.88
0.0115	34.0	1938	0.9989	0.88
0.0001	35.0	1995	0.9871	0.88
0.0001	36.0	2052	0.9920	0.88
0.0002	37.0	2109	0.9974	0.88
0.0002	38.0	2166	0.9950	0.88
0.0001	39.0	2223	0.9997	0.88
0.0001	40.0	2280	0.9951	0.88

Framework versions

Transformers 4.31.0
Pytorch 2.0.1+cu117
Datasets 2.13.1
Tokenizers 0.13.3

📄 License

This model 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