The following project is based on works from the Prado Museum:
- Identify all the existing faces from the art.
- Encode and store the faces in embeddings.
- Upload a portrait and get the closest embedding.
By following these steps, the user will be able to see who he/she looks like in the Prado Museum or even find your Doppelgänger!
The project is divided into two parts: an API that encodes an endpoint to search for the most and search for the most similar face in the faces directory, and another part to interact with the user in uploading and displaying the results, under the web directory.
├── assets <- Miscellaneous folder, images from the repository
|
├── faces <- Identification and encoding of Prado faces, API service
│
└── web <- Web interface to search a nearest face
Each directory contains its own README documenting that specific part, I encourage you to explore them if you need more information.
The dataset is available in Kaggle where we can access it freely and download it.
Once the services have been launched, it is necessary to define a recurring task to clean the directory where user photos are uploaded. This is done using Crontab. We can configure it as follows:
crontab -eOne with the editor of your choice, add the following line at the end of the file:
*/5 * * * * rm -rf /tmp/dobble/*