project-capstone-zoomcamp

Problem Description

classify whether the person in the photo is wearing glasses or not .Dataset from https://www.kaggle.com/aniruddha123/glasses-data . I make two deep learning model ,model_v1 and model_v2.The difference just in number of convolutional layer.

File in repository :

• model_v1.ipynb : Model selection process
• model_v2.ipynb :
• predict-test.py : to try web service that deployed locally
• deploy-test.py : to try web service that deployed on pythonanywhere
• Dockerfile : to running the service on docker
• model_v1_10_0.948.h5 : Model that i save from training
• model_v1_10_0.982.h5 : Model that i save from training
• requirements.txt : requirement to run model_v1.ipynb or model_v2.ipynb
• make_folder.ipynb : To make train,validation,and test folder

For EDA , i referenced from this source :

• https://medium.com/geekculture/eda-for-image-classification-dcada9f2567a
• https://towardsdatascience.com/exploratory-data-analysis-ideas-for-image-classification-d3fc6bbfb2d2

How to run project

Install dependencies

1. Create new environment with conda create -n myenv python=3.8
2. Activate new environment with conda activate myenv
3. Install dependencies with pip install - r requirements.txt

To run notebook.ipynb

1. Activate myenv
2. Open model_v1.ipynb or model_v2.ipynb . I put EDA and problem description in model_v1.ipynb

Deploy locally using docker

1. Build docker image using docker build -t glass-model .
2. run docker image using docker run -it --rm -p 8080:8080 glass-model:latest
3. run python predict-test.py on another command prompt

Thank you for reviewing my project , sorry for the many shortcomings