Basic usage

Loading the tools

compIAM does not have terminal functionalities but it is to be used within Python based-projects. First, import the library to your Python project with: import compiam.

The integrated tools and models are organized by:

  1. First, the following fundamental musical aspects: melody, rhythm, structure, timbre, and from v0.3.0 we have included separation as well.

  2. Then, the tools are grouped by tasks.

You can access the several included tools by importing them from their corresponding modules:

from compiam.melody.pitch_extraction import FTANetCarnatic
from compiam.rhythm.transcription import FourWayTabla

We provide nice functionalities to explore where do the tools live:

  1. Print out the available tasks for each category: compiam.melody.list_tasks()

  2. Print out the available tools for each module using: compiam.melody.list_tools()

  3. Print out only the tools for a particular task: compiam.melody.pitch_extraction.list_tools()

Important

Some tools (especially the ML/DL models) require specific dependencies that are not installed by default, because of their size or compatibility issues. If a tool is loaded and a particular dependency is missing, an alert will be displayed, to inform the user which dependency is missing and how to proceed to install it in the right version. See optional_requirements.txt where the optional dependencies and the specific versions we use in compiam are listed.

Wrappers

compIAM also includes wrappers to easily initialize relevant datasets, corpora, and also pre-trained models for particular problems.

compiam.load_dataset(dataset_name[, ...])

Alias function to load a mirdata Dataset class.

compiam.load_corpora(tradition[, cc, token])

Wrapper to load access to the Dunya corpora.

compiam.load_model(model_name[, data_home, ...])

Wrapper for loading pre-trained models.

Tip

When listing available tools using the list_tools() functions, some will appear with a “*” at the end. That is meant to indicate that such tools have pre-trained models available, which may be loaded using the wrapper compiam.load_model().

Model weights are large in size and therefore, not included in the library from v0.2.1 on. We have included a .download_model() function to all ML/DL models that require pre-trained weights, so that the user can download them on demand. This function is automatically run when invoking the model through the compiam.load_model() wrapper. The model weights are then stored in the corresponding default folder ./compiam/model/. If the model is already downloaded, the function will not download it again.

Note

From v0.3.0 on, compiam.load_model() wrapper has an argument data_home, in where you can specify to which folder you want the models to be downloaded and read from.