aekasitt/kaggle-chula-utc-tale-of-two-wats

The Tale of Two Wat

For Chulalongkorn AI Academy candidate screening

• Compatible with Python 3.6 and 3.7

Problem introduction

For this problem, you are tasked to develop a machine learning model that takes a photo as input and classify whether that photo contains Wat Prakaew or Wat Po. Gather your dataset from photos on the internet.

Getting Started

Run the following command in your Terminal to install Project Dependencies

python install -r requirements.txt

This project requires the collection of owned data, run the following command to begin scraping Wat Pho and Wat Prakeaw Images from Google Image Search using FireFox Webdriver Please download and place geckodriver on root-folder

python scrape.py "wat po" data/watpo --amount 500
python scrape.py "wat prakeaw" data/watprakeaw --amount 500

Also insert test images downloaded from Kaggle dataset to the following folders:

data/watpo/test data/watprakeaw/test

Code Walkthrough

Please open image_recogntion.html (or image_recogniton.ipynb if you have Jupyter Notebook instaalled) and follow the code walk-through there. The project creates and saves a convolution neural network estimator for image processing, saved under pickles/convulution_model.pkl

Sample Accuracy of create model tested against 8 images found on Kaggle Competition Page:

[INFO] Floating Point Errors:
[INFO] R2: -0.4521
[INFO] Mean Squared Error: 0.3630
[INFO] Root Mean Squared Error: 0.6025
[INFO] Binary Classification Errors:
[INFO] R2: -0.5000
[INFO] Mean Squared Error: 0.3750
[INFO] Root Mean Squared Error: 0.6124