Cyrus Asfa-Vahidi's Projects
This sketch combines uses a long term average, a short term average and the summation of the delta between those two to detect beats in music. An FFT splits apart the frequency bands, from that the desired band is isolated and processed. The short term average is compared to the long term average, the the use of a threshold to detect beats. A multiplier is calculated based on the overall volume, and the summation of the delta between the two averages over a relatively long period of time. This adjusts for the different styles of music that require different thresholds.
Perceptual Musical Similarity Metric Learning with Graph Neural Networks - WASPAA 2023
An introduction to python for data analysis (for MAT students)
A real-time sample granulation implementation run on Bela, ARM Cortex-A8
tutorial on 1D wavelet scattering (Kymatio Workshop 2022)
A Implementation of SpecAugment with Tensorflow & Pytorch, introduced by Google Brain
SuperCollider, SuperDirt and TidalCycles startup files, configs and functions
Synthesizer plug-in (previously released as Vember Audio Surge)
subtractive synth data generation in supercollider
Tape Delay based on Roland RE-201 Space Echo and Echoplex Tape Echo - on JUCE
Fast Ethereum RPC client for testing and development
A collection of metrics for evaluating timbre dissimilarity using the TorchMetrics API
Dissimilarity Matrix and Sounds from Timbre Space Representation of a Subtractive Synthesizer (Timbre, 2020)
Online Triplet Loss - Pytorch Lightning - MNIST
A GPU-optional modular synthesizer in pytorch, 16200x faster than realtime, for audio ML researchers.
Implementation/Interpretation of Vector Quantized Timbre Representation (Bitton et al 2020)
wacc language compiler (imperial second year project)