One of my side hobbies is the purchasing and selling of musical instruments. Over the past few year, I have built automation tools to identify musical instruments being sold under market value. In order for my bots to identify musical instruments under market price, I need a computer vision models to determine the value and origin of musical instruments from only one photo. I therefore created a database of over 10,000 instruments with maker name, year built, location built, auction data, and photos of the instrument through an extensive web scraping project. Included in my project is an interactive GUI guessing game, data visualizations of the violin trade, an inflation estimator for stringed instruments, a website, and market analysis of the violin auctions. Below is my GitHub repository showing some, but not all, of my efforts on this front.

Below is a PowerPoint of a presentation I made showing the power of transfer learning for musical instrument classification.

Using Transfer Learning for Musical Instrument Classification

Here is the associated paper reviewing the motivation and details for using transfer learning to classify musical instruments.  

Using Transfer Learning for Musical Instrument Classification.pdf

History of The Project

I started with creating a database of thousands of instruments with maker names, country built, location made, and year made, along with photos of the instrument (2020).

Then I built convolutional neural networks that can predict an instrument’s price, age, and location built.

I then built An instrument guessing game which taught me how to predict the maker name, year made, price, and origin of musical instruments.

A history of over 400 years of violin making has led to the existence of billions of dollars worth of instruments. I wrote code to assemble 150 million dollars worth of instruments into a pandas database. I made data visualizations of the violin trade. I made an inflation estimator for all instruments. I designed and implemented a convolutional neural network to predict the worth of an instrument, and most recently, a GUI to guess musical instruments. Included: Visualizations of Spread of Violins over time