Celebrity Look Alike Finder With Python DeepFace Module

The Python Deepface module has recently become popular for its ease in creating machine learning projects. So I thought of giving it a try. Forgive me if I mistake somewhere, I am still learning

So I went to kaggle.com and got a Celebrity Face Image Dataset which contains 100 images of different styles of each of the following actors and actresses:

• Angelina Jolie
• Brad Pitt
• Denzel Washington
• Hugh Jackman
• Jennifer Lawrence
• Johnny Depp
• Kate Winslet
• Leonardo DiCaprio
• Megan Fox
• Natalie Portman
• Nicole Kidman
• Robert Downey Jr.
• Sandra Bullock
• Scarlett Johansson
• Tom Cruise
• Tom Hanks
• Will Smith

The Code

We use Jupyter notebook for this project.
The first snippet goes as

import os
import shutil

data_dir = 'Celebrity Faces Dataset'

for directory in os.listdir(data_dir):
    first_file = os.listdir(os.path.join(data_dir, directory))[1]
    shutil.copyfile(os.path.join(data_dir, directory, first_file), os.path.join("Samples", f"{directory}.jpg"))

What it does is it extracts the first image from the folders of every celebrity and copies it and pastes it in a separate folder called Samples by the name of the actor/actress. This will later be used the the ML model to detect matches.

The second and the final snippet is:

from deepface import DeepFace

smallest_distance = None

for file in os.listdir("Samples"):
    if file.endswith('.jpg'):
        result = DeepFace.verify('me1.jpg', f'Samples/{file}')
        if result['verified']:
            print("This person looks exactly like", file.split('.')[0])
            break
        if smallest_distance is None:
            smallest_distance = (file.split('.')[0], result['distance'])
        else:
            smallest_distance = (file.split('.')[0], result['distance']) if result['distance'] < smallest_distance[1] else smallest_distance
else:
    print(f"No exact match found. Closest match is {smallest_distance[0]}")

Here we import the Deepface module. The Deepface module has a function called DeepFace.verify() which verifies an image pair is same person or different persons. In the background, verification function represents facial images as vectors and then calculates the similarity between those vectors. Vectors of same person images should have more similarity (or less distance) than vectors of different persons.

The output is stored in a variable called result.
If the supplied image exactly matches with one of the images of the Dataset, i.e, the result['verified'] parameter gets equivalent to True the program responds with a message showing the celebrity the supplied subject resembles.

The smallest_distance variable is assigned None initially. It later is assigned the value of the result['distance'] which is a numeric value showing the closest match of the subject to the data and described by the module author as distance_metric (string): cosine, euclidean, euclidean_l2. The program verifies if the smallest_distance variable is None in every iteration through the samples. For not being None, if the smallest_distance value is lesser than the value of result['distance'] from the latest iteration, the new result['distance'] is assigned to the smallest_distance variable. To note that the smallest_distance variable contains the name of the celebrity too since it is stored as a string separated with a . where the 0th index holds the celebrity name in the current iteration while the 1st index holds the value of result['distance'] of that iteration. At the end when the loop ends and no exact match is found, the last most closest match iterated is retrieved and printed in the console to the user.

Some notes from my experience

I personally saw that the model is not totally accurate, in fact it matched one of my pics to Angelina Jolie once (I am male). This greatly depends on the image passed. If no face is in the given image, the program will raise error.

You can view the code here: (https://github.com/Sannidhya127/Celebrity-Look-Alike-Finder)

I would really appreciate a star on the GitHub repo. Thank you and have a great day!

Comments

[Aswet Natex]

Celebrity gossip offers a fun and engaging way to stay connected with the lives of your favorite stars, from red carpet events to personal milestones. It provides fans with a glimpse into the glamour and excitement of the entertainment world, sparking conversations and bringing people together over shared interests. While it’s important to approach it with respect for privacy, celebrity gossip can also inspire through stories of success, resilience, and creativity, making it a lighthearted escape and a source of entertainment for many.