udacity-ml-capstone

Introduction

For my Udacity Machine Learning Capstone project I chose to implement a Fruit classifier that predicts the label of the input frut image.

Reports

• report.pdf
• proposal.pdf

Code

• 1- Data Exploration & Basic CNN.ipynb : has some visualization of data and its distribution, also basic CNN model

• 2- Tuning CNN.ipynb : has implementation of 3 different tuning for basic CNN model

• 3- Tuned CNN with Image Augmentation.ipynb : has image augmentation with one of the tuned CNN models

• 4- Transfer Learning with Image Augmentation.ipynb : has two implementation of transfer learning one with pretrained weights and another one using random weights(both are Xception models)

Note that the code is run on Kaggle not locally on jupyter notebook

Input files

• Main Dataset(for test and training) can be downloaded from here (https://www.kaggle.com/moltean/fruits/kernels)
• Real Test Set that i gathered my self (https://www.kaggle.com/yahiaelshahawy/realtestdata)
• Keras pretrained weights for working on kaggle can be downloaded from here (https://www.kaggle.com/gaborfodor/keras-pretrained-models)

Technologies used

• Python 3.x
• Tensorflow
• Keras