Music Classification and Recommendation using CRNN

• Tech Stack: Python, Keras, Tensorflow, Librosa
• Github URL: Project Link

Pioneered genre classification using the Million Song Dataset. Streamlined raw data preprocessing and harnessed the power of the Librosa library for feature extraction.

Generated mel-spectrogram images from extracted features via Keras, followed by training of a custom-tailored CRNN model with optimized hyperparameters. Calculating multidimensional distances between distinctive feature vectors for each song, the system recommends 10 analogous tracks.