victor369basu/audio-track-separation

In this Repository, We developed an audio track separator in tensorflow that successfully separates Vocals and Drums from an input audio song track.

Hi there 👋

I am Victor Basu, I'm a passionate Data Scientist and Machine Learning Engineer with a strong background in turning data into actionable insights and building intelligent systems. My journey in the world of data and machine learning began with a curiosity to unravel the hidden patterns in data and use them to make informed decisions. I am a kaggle Notebooks Master, you could follow me on Kaggle at @basu369victor. I have a hand full of experience with the technologies required today at the industry level. Other than Data Science and Machine Learning I do take some interest in web-development stuffs. You could also follow me on LinkedIn at @Victor Basu

My contribution to Keras -

• Drug Molecule Generation with VAE
• An approach towards implementing a depth estimation model with a convnet.

My work on Attention based Protein Structure Prediction

Highlight

Facial Emotion Recognition

In this project I have developed an end-to-end pipeline for real-time Facial emotion recognition application through full-stack development. The frontend is developed in react.js and the backend is developed in FastAPI. The emotion prediction model is built with Tensorflow Keras, and for real-time face detection with animation on the frontend, Tensorflow.js have been used.

GitHub project repository - Facial Emotion Recognition

High quality youtube video available at - https://youtu.be/aTe05n6T5Vo

Architecture to host QuickSight Dashboard for HuggingFace model monitoring deployed on SageMaker along with data EDA

This is a solution that demonstrates how to train and deploy a pre-trained Huggingface model on AWS SageMaker and publish an AWS QuickSight Dashboard that visualizes the model performance over the validation dataset and Exploratory Data Analysis for the pre-processed training dataset. With this as the architecture for the proposing solution, we try to solve the classification of medical transcripts through Machine Learning, which is basically solving a Bio-medical NLP problem. In this solution, we also discuss feature engineering and handling imbalanced datasets through class weights while training by writing a custom Huggingface trainer in PyTorch.

GitHub project repository - Host QuickSight Dashboard for HuggingFace model monitoring deployed on SageMaker along with data EDA Demo video - https://youtu.be/RhTSnn41cnM

audio-track-separation's Issues

About Separating Track

Hi,

How can I separate .wav file to vocal and instrumental(accompaniment) .wav files? I can't see it in output folder.

checkpoint model request

Hi, very excited about this tool that looks promising,

But I would like to make a request, would it be possible to release the pre-trained model? to test on our own songs,

We are already grateful.