Custom Image Classifier Using Keras Tuner

Building an image classification model using keras tuner with customize datasets can be efficient and effective appraoch for developing optimized architecture for CNN model.

There are five different music symbols considered for the classification.

1. Whole Note
2. Half Note
3. Quarter Note
4. Eight Note
5. Sixteenth Note

Datasets Format:

There are 5000 Grayscale image (28x28 pixels) samples of 5 different music note symbol used. data format is similar to the MNIST datasets.

Major Python-Libraries used:

1. tensorflow 2.3.0
2. keras-tuner: 1.0.1
3. matplotlib: 3.2.2
4. pandas: 1.0.5
5. numpy: 1.18.5

References:

https://www.kaggle.com/kishanj/music-notes-datasets

https://keras-team.github.io/keras-tuner/

https://blog.tensorflow.org/2020/01/hyperparameter-tuning-with-keras-tuner.html

https://keras-team.github.io/keras-tuner/documentation/tuners/