The goal of this project is to utilize advanced image processing and machine learning techniques to predict the attractiveness of a face in an image. This prediction is based on several visual features derived from the face, enabling automated attractiveness classification.
Data Exploration and Preprocessing
To begin with, an extensive exploratory data analysis (EDA) was conducted on the celeba dataset on kaggle, allowing me to better understand the relationship between the features and the target variable (Attractiveness). The dataset included face images along with their respective attractiveness scores.
Model Building
I worked with several deep learning models to tackle the classification problem. These models include:
- InceptionV3
- ResNet-50
- ResNet-101V2
Each model was fine-tuned by freezing certain layers and adding custom fully connected layers to adapt them for the task at hand.
To ensure that the model can handle various transformations in images such as rotation, flipping, and scaling, I implemented data augmentation. This helped in improving the model's generalization capabilities and performance across diverse face images.
The models were trained using a subset of the dataset, which was split into training, validation, and testing sets. Performance metrics such as accuracy were used to evaluate the models, and the best-performing model was selected.
Performance
The project achieved the following accuracy scores for the respective models:
- InceptionV3: 86.80%
- ResNet-50: 70.20%
- ResNet-101V2: 72.13%
The InceptionV3 model achieved the highest accuracy, demonstrating its ability to effectively classify faces based on attractiveness.
Application Building
Streamlit application is build into 2 steps:
- Step 1: Recognize the face from the imported media either image or video or live camera.
- Step 2: Transfer that face for pre-processing and prediction to the model pipeline.
This project highlights my proficiency in deep learning, model deployment, and web application development, making it a valuable addition to my portfolio.