Face search by photo
About project
The user needs to find the names of actors and films with their participation by photo.
A task
Now the user can find films with the participation of actors who are shown in the photos. Users spend more time on the site. Advertising revenue has increased.
Business effect
The appearance of a function allows the user to find the names of actors in the photo. Now the user can find movies with these actors.
Solution
Customer Journey Map
How does the service work?
A process that includes detection, alignment and vectorization faces using the DeepFace library.
Facial analysis is converted into a digital code.
Search with E-lasticsearch. E-lasticsearch is the base data with built-in algorithm HNSW from the ANN family, thanks to which, there is a quick search.
Uploading a photo with an actor to the site
1
Face-recognition and analysis
The results: the name of the desired actor and films with him
Data conversion
2
Search results
Teсhnology
A framework for creating web applications, thanks to which the logic of interaction with a neural network via the Internet is configured.
Flask
Converts requests received from Nginx into a format that the web application can process, and also runs code when necessary.
Gunicorn
A resource for demonstrating the visual component after training a neural network. Thanks to Streamlit, the customer can test the neural network before launching the service. So the product can be improved if there are bad results.
Streamlit
Thanks to this function, the model learns to separate figures in such a way as to find the right person among similar ones.
ArcFace
Docker is a tool that has all the necessary libraries and programs for the operation of the neural network. Thanks to him, the neural network can be easily and easily launched on any computer.
Downloads and launches the necessary services from the Internet. It can process them very quickly and is usually configured to start only those services that the web application really needs.
Nginx
Elasticsearch is a database with a built-in ANN algorithm not found in other databases.
ANN (Approximate Nearest Neighbor) is an algorithm that quickly and efficiently finds a huge number of examples.
ANN (Approximate Nearest Neighbor) is an algorithm that quickly and efficiently finds a huge number of examples.
Elasticsearch
Docker
Project team
Frank Sh.
Project Management / Machine learning
Dmitry I.
Machine learning /
Data analysis and technology
Data analysis and technology
Areas of use
This technology can be used to recognize fraudsters who are trying to impersonate another person using fake documents. Use if the company has a database with photos and information about its customers.
