Evaluation Framework for CCMusic Database Classification Tasks

Classify spectrograms by fine-tuned pre-trained CNN models.

Download

git clone [email protected]:monetjoe/ccmusic_eval.git
cd ccmusic_eval

Environment

conda create -n py311 python=3.11 -y
conda activate py311
pip install -r requirements.txt

Usage

python train.py --ds ccmusic-database/bel_canto --subset eval --data cqt --label singing_method --model squeezenet1_1 --wce True --mode 0

Help

Args	Notes	Options	Type
--ds	The dataset on ModelScope to be evaluated	For examples: ccmusic-database/CNPM, ccmusic-database/bel_canto	string
--subset	The subset of the dataset	For examples: default, eval	string
--data	Input data colum of the dataset	For examples: mel, cqt, chroma	string
--label	Label colum of the dataset	For examples: label, singing_method, gender	string
--model	Select a CV backbone to train	Supported backbones	string
--imgnet	ImageNet version the backbone was pretrained on	v1, v2	string
--mode	Training mode ID	0=linear_probe, 1=full_finetune, 2=no_pretrain	int
--bsz	Batch size	For examples: 1, 2, 4, 8, 16, 32, 64, 128..., default is 4	int
--eps	Epoch number	Default is 40	int
--wce	Whether to use weighted cross entropy	False, True	bool

Fixed Hyper Params

Param	Value	Range
iteration	10	train
lr	0.001	optimizer
momentum	0.9	optimizer
optimizer	SGD	scheduler
mode	min	scheduler
factor	0.1	scheduler
patience	5	scheduler
verbose	True	scheduler
threshold	lr	scheduler
threshold_mode	rel	scheduler
cooldown	0	scheduler
min_lr	0	scheduler
eps	1e-08	scheduler

Cite

@dataset{zhaorui_liu_2021_5676893,
  author       = {Monan Zhou, Shenyang Xu, Zhaorui Liu, Zhaowen Wang, Feng Yu, Wei Li and Baoqiang Han},
  title        = {CCMusic: an Open and Diverse Database for Chinese and General Music Information Retrieval Research},
  month        = {mar},
  year         = {2024},
  publisher    = {HuggingFace},
  version      = {1.2},
  url          = {https://huggingface.co/ccmusic-database}
}