Seam Carving as a Data Augmentation Technique
For more detail please check out my website
Also on pipy
A Python library for performing seam carving on audio files, converting audio to matrix and back, and generating spectrograms.
You can install the library using pip:
pip install AudioCarver==0.1.0Example in a python file:
from audio_carver import carve_audio, sig_to_wav
# Carving in time domain and writing to output
magnitude, phase = carve_audio(number_of_seams, magnitude, phase, is_vertical=True) # default true
sig_to_wav('output.wav', magnitude, phase)Example in terminal(Recommended):
$ python main.py <input_file.wav> <number_of_seams> <--carve_time> # remove to carve in frequencyCurrent Supported Audio Formats: .wav
Audio Processing Capabilities: Spectral analysis, Time-domain operations, Signal modification
Input Requirements: None
Output: Mono signal with sampling rate inherited from the original input file's sampling rate
VIsualization: Spectrogram comparisons of before/after carving
How to load an audio file:
from audio_carver import *
audio, sampling_rate = load_wavfile('filename.wav')Find the minimum seam:
from audio_carver import *
minimum_energy_seam = min_vertical_seam_energy(magnitude)Documentaion will be uploaded soon.
This project is licensed under the MIT License - see the LICENSE file for details.
Please contact alicjam@uw.edu for questions, suggestions, etc. Thank You!