Sports Celebrity Image Classifier

In this data science and machine learning project, I have classified sports personalities. I have restricted classification to only 5 people,

1. Maria Sharapova
2. Serena Williams
3. Virat Kohli
4. Roger Federer
5. Lionel Messi

Technologies used in this project

1. Python 3.8
2. Numpy and OpenCV for data cleaning
3. Matplotlib & Seaborn for data visualization
4. Sklearn for model building
5. Jupyter notebook, visual studio code and pycharm as IDE
6. Python flask for http server
7. HTML/CSS/Javascript for UI

Folder structure

• UI : This contains ui website code
• server: Contains the Python flask server related code
• model: Contains python notebook for model building
• google_image_scrapping: Contains the code to scrap google for images
• images_dataset: Dataset used for training our model

Model Selection and Methodology

In this project, OpenCV haarcascade was used to detect face and two eyes. Wavelet transform was used to extract these features from the images. Finally after processing and cleaning the data, it was given as input to models. Various models were tried using GridSearchCV and Logistic Regression was found to give the best resutls with a score of 84.15%.


Above is the heatmap showing the results of Logistic regression on test data. For test images, Logistic regression was found to give a score of 95%. The model accuracy could be improved by using a bigger dataset. Also deep learning may give better results for image classificaion