This last month, I have submitted my doctoral thesis entitled “Deep Neural Networks for Music and Audio Tagging”. I’ll be defending next November 15th, and I’m excited to announce that my jury will be conformed by Geoffroy Peeters, Perfecto Herrera, and Juhan Nam.
The musicnn library (pronounced as “musician”) employs deep convolutional neural networks to automatically tag songs, and the models that are included achieve the best scores in public evaluation benchmarks. These state-of-the-art models have been released as an open-source library that can be easily installed and used. For example, you can use musicnn to tag this emblematic song from Muddy Waters — and it will predominantly tag it as blues!
As part of my onboarding at Dolby, I had the pleasure to be working in San Francisco. In order to share my recent experiences with my colleagues, I have been updating these slides and I presented some of my recent work at Dolby and Adobe headquarters.
We present a didactic toolkit to rapidly prototype audio classifiers with pre-trained Tensorlow models and Scikit-learn. We use pre-trained Tensorflow models as audio feature extractors, and Scikit-learn classifiers are employed to rapidly prototype competent audio classifiers that can be trained on a CPU.
This material was prepared for teaching Tensorflow, Scikit-learn, and deep learning in general. Besides, due to the simplicity of Scikit-learn, this toolkit can be employed to easily build proof-of-concept models with your own data.
During the last summer, I have been a research intern at Telefónica Research (Barcelona). The article “Training neural audio classifiers with few data” is the outcome of this short (but intense!) collaboration with Joan Serrà, where we explored how to train deep learning models with just 1, 2 or 10 audios per class. Check it out on arXiv, and reproduce our results running our code! These slides are the extended version of what I will be presenting next week in ICASSP! See you in Brighton 🙂