Home

a.k.a.

Realtime Audio Classification for Musicians

(As an homage to Tensorflow for Poets.)

Overview

In this workshop, we will teach you how to design audio classifiers using neural nets. We will guide you through the steps of collecting and organizing data, generating spectrographs, training a network, and the using that network to detect audio in realtime. We will use Jupyter notebooks, Python3, pyTorch, and Librosa to play with neural nets that can detect different music and different audio sources.

Provisional Workshop Schedule

Time	Monday	Tuesday	Wednesday	Thursday	Friday
9am	Introductions	Review/Q&A	Review/Q&A	Review/Q&A	Review/Q&A
10-noon	Neural Nets	Collecting & Analyzing Sounds	Designing Interaction	Generative Models	Project time
noon-1:30	Lunch	Lunch	Lunch	Lunch	Lunch
1:30-3:30	Lab Setup	Stanford Sounds Dataset Activity	Urban Sounds Dataset Lab	Final Project	Project Time/ Show and Tell
3:30-5pm	Cats & Dogs Lab	Stanford Sounds Dataset Activity	Plotting Final Project	Final Project	Happy Hour

Labs

Lab 0. Setting up

Lab 1. Cats and Dogs--binary classification

Lab 2. Making the Stanford Sounds Dataset

Lab 3. Urban Sounds Dataset

Lab 4. Making stuff interactive

Lecture Slides

Introduction

The Bitter Lesson