Completing Data Collection and Cleansing, as well as with resize data onto same size

Completing Model Buiding and Tuning

Completing Model Deployment Setup

• Data Collection has done using web scraping method through Google Images.

• Data Cleansing has done on images that is not relevant with the real traditional music instruments.

• Data Resize has done by converting various different size of images onto the same size of 256x256 px.

• Data Detail:

  • Bonang: 127 images
  • Kolintang: 101 images
  • Rebab: 68 images
  • Saluang: 107 images
  • Sape: 147 images
  • Sasando: 98 images
  • Tifa: 88 images

• Model has been builded using custom model and adapted by our own datasets and InceptionV3-based model architecture.

• Model have better performance and precision than the previous model by converting all training images into grayscale

• Model is saved in HDF5 format

• Model is deployed with Flask

• Model is deployed in Google Cloud Platform

• Model is deployed using CI/CD cycle and Docker