Project Description

In this project, we take on the exciting task of building a facial recognition system that can identify Arnold Schwarzenegger in a crowd of faces.

We are pretending to be part of a top-level data science team that creates advanced AI tools to protect famous people—like actors, athletes, politicians, and philanthropists—by recognizing their faces using machine learning.

For this project, our focus is Arnold Schwarzenegger, a well-known public figure who has been a bodybuilder, movie star, and governor.

---

Goal

Build a machine learning system that can:

• Distinguish between images of Arnold Schwarzenegger and others.
• Compare three different classification models.
• Choose the best model based on cross-validation performance.

---

The Dataset

The dataset is called /lfw_arnie_nonarnie.csv and is based on facial images from the Labeled Faces in the Wild (LFW) dataset.

It contains:

• 40 images of Arnold Schwarzenegger.
• 150 images of other people.

Columns:

Column Name	Description
PC1, PC2, ..., PCN	These are principal components (from PCA), which represent important features of the face images.
Label	1 = Arnold Schwarzenegger 0 = Someone else

---

What We’ll Do

1. Build ML pipelines for three different models (e.g., Logistic Regression, Random Forest, SVM).
2. Use cross-validation to evaluate how well each model performs.
3. Compare the models and select the one with the highest accuracy or best score.

---

Final Outcome

Create a facial recognition model that can reliably identify Arnold Schwarzenegger, helping enhance the safety and recognition of high-profile individuals using machine learning.

# Import required libraries
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import GridSearchCV, KFold, train_test_split
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, confusion_matrix

# Read the CSV file 
df = pd.read_csv("lfw_arnie_nonarnie.csv")
df.head()

	0	1	2	3	4	5	6	7	8	9	...	141	142	143	144	145	146	147	148	149	Label
0	-2.061987	0.581320	-0.249115	-0.631340	-1.359899	0.751619	-0.029364	1.610456	0.341402	1.286709	...	-0.356844	-0.016488	-0.228473	0.258134	0.046834	0.135742	-0.068297	0.022038	0.090003	1
1	-0.796838	-0.667228	-0.107889	0.019755	-0.686348	0.912779	0.463412	-0.238308	-0.294023	0.215470	...	-0.037243	-0.012105	-0.351285	-0.034968	0.192314	-0.015406	-0.089117	0.023588	-0.019998	1
2	5.376779	1.142695	2.543111	-2.727212	0.272785	-0.972187	1.111221	1.645502	-2.556968	-0.648781	...	0.157441	-0.333875	-0.303720	-0.085975	0.171346	0.128577	-0.118262	0.045881	-0.190158	1
3	7.029235	1.242883	-2.628079	1.224479	-1.141370	-1.620647	0.205890	1.567561	0.736200	0.010782	...	0.051040	-0.068796	0.141841	-0.227999	0.046044	0.013643	-0.125893	0.146396	0.013320	1
4	5.484822	6.752706	-4.291114	1.740412	-1.603087	-1.075175	1.919936	-0.197615	1.030596	1.451936	...	0.034412	0.265141	0.226000	0.032064	-0.113654	0.059126	-0.216803	0.025849	0.020456	1

5 rows × 151 columns

# Seperate the predictor and class label
X = df.drop('Label', axis=1)
y = df['Label']

# Split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=21, stratify=y)

# Store initialized models in a dictionary
models = {"LogisticRegression": LogisticRegression(), 
          "KNeighborsClassifier": KNeighborsClassifier(),
          "DecisionTreeClassifier": DecisionTreeClassifier()}



# Store the model parameters in a dictionary
param_grid = {"LogisticRegression": {"LogisticRegression__C": [0.01, 0.1, 1, 10]},
              "KNeighborsClassifier": {"KNeighborsClassifier__n_neighbors": range(1,10)},
              "DecisionTreeClassifier": {"DecisionTreeClassifier__max_depth": [2, 5, 10],
           "DecisionTreeClassifier__min_samples_split": [2, 5, 10, 20],
           "DecisionTreeClassifier__random_state": [42]}}



# Define cross-validation parameters
kf = KFold(n_splits=5, random_state=42, shuffle=True)



# Prepare to collect Grid Search CV results
pipe_accuracies = {}
pipe_params = {}
pipelines = {}

# Create separate pipelines for each model, loop through the models and perform GridSearchCV
for name, model in models.items():
    pipeline = Pipeline(steps=[
        ("scaler", StandardScaler()),
        (name, model)
    ])
    # Create the GridSearchCV object
    grid_search = GridSearchCV(pipeline, param_grid[name], cv=kf, scoring="accuracy")
    
    # Perform grid search and fit the model and store the results
    grid_search.fit(X_train, y_train)
    pipe_accuracies[name] = grid_search.best_score_
    pipe_params[name] = grid_search.best_params_
    pipelines[name] = grid_search

# Select the best model based on the best cross-validation score
best_model_name = max(pipe_accuracies)
best_model_cv_score = max(pipe_accuracies.values())
best_model_info = pipe_params[best_model_name]

# Print the best model name, parameters, and CV score
print(f"Best Model: {best_model_name}")
print(f"Best Model Parameters: {best_model_info}")
print(f"Best Model CV Score: {best_model_cv_score}")

Best Model: LogisticRegression
Best Model Parameters: {'LogisticRegression__C': 1}
Best Model CV Score: 0.8288172043010752

# Compute and print key performance metrics
y_pred = pipelines[best_model_name].predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1 = f1_score(y_test, y_pred)

print(f"Accuracy: {accuracy:.4f}")
print(f"Precision: {precision:.4f}")
print(f"Recall: {recall:.4f}")
print(f"F1 Score: {f1:.4f}")

Accuracy: 0.8158
Precision: 1.0000
Recall: 0.1250
F1 Score: 0.2222

The best performing model based on cross-validation scores is LogisticRegression with accuracy of 0.8158