Add Music Accompaniment Generator Project to ML-Nexus #121
Conversation
|
Thank you for submitting your pull request! 🙌 We'll review it as soon as possible. In the meantime, please ensure that your changes align with our CONTRIBUTING.md. If there are any specific instructions or feedback regarding your PR, we'll provide them here. Thanks again for your contribution! 😊 |
|
Where did you raised an issue?...it's not the place to to raise an issue it's the place to make PR's for raised and assigned issues. |
I have raise the issue in https://github.com/UppuluriKalyani/ML-Nexus/issues issue #122 , and here is my solution to the issue. I hope my project can be merged to the main branch. |
|
@NanFangDieDao can you please add readme file with clear description and with attached results..and also show here code works(results) it could be screen recording or screen shots...but I would prefer screen recording |
I've just added a README document to my project, and I believe I've described my project and its results in sufficient detail within it. @UppuluriKalyani |
Melodic accompaniment is an indispensable part of the music creation process. Traditionally, it has a high professional threshold. This undoubtedly poses a limitation to people lacking deep background knowledge of music who are motivated to undertake this activity for a variety of reasons, such as interest. A solution is needed to simplify the task of melodic accompaniment and enable non-specialists to participate in simple melodic accompaniment activities. At the same time,
Transformer, which has emerged in the field of deep learning in recent years, has demonstrated high performance and great potential in various fields, and therefore, the use of deep learning techniques based on the Transformer model to simplify the task of melodic accompaniment is of great research value and possible to realize.
I use Python and Pytorch framework to design and implement the Transformer-based melody automatic accompaniment model.
In this project,I implement the innovative rule-based melody recognition algorithm and rule-based alignment algorithm to preprocess the dataset.After that I construct the Transformer model, including the embedding layer, the positional embedding layer, the self-attention layer, and encoder,decoder structures. The model can accept a MIDI file of the main melody and generate a corresponding MIDI file including main melody track together with other accompaniment instrument tracks such as piano,guitar,bass and drum.The project also provides the training, evaluation and validation results of the model.