Project Overview
- Designed and implemented a high-performance deep learning pipeline to classify music genres from raw audio data
- Transformed audio signals into spectrogram representations, enabling convolutional and recurrent neural networks to learn spatial–temporal patterns
- Built and evaluated CNN, RNN, and LSTM architectures, comparing accuracy, training time, and computational efficiency
- Applied parallel computing strategies to accelerate model training and improve scalability over traditional sequential approaches
- Analyzed performance trade-offs between model complexity and inference efficiency for real-time and streaming use cases
- Demonstrated system-level ML thinking, optimizing both model accuracy and computational performance
Repository
🎵 High-Performance Scientific Compute
Deep learning pipeline for music genre classification using CNN, RNN, and LSTM on spectrograms with parallel computing for accelerated training.
View on GitHub →