Why create a Face Library?
Creating a Face Library enables you to build a database of faces. This database is used to recognize people in media files quickly and efficiently and search across your media for those people.
How to Create a Face Entity
1. Navigate to the Library App through the App Switcher. The app will open in a new tab.
2. Select the Library to which you’d like to add your new person, or Entity, by clicking on the Library name on the My Libraries page:
3. This will open your Library. At the top-right corner of the page, click Add Person:
4. This will open the Add New Person Console. Enter the name of the person in the designated field at the top. Then click on Edit within the avatar circle to add a cover image. Add a Description if desired. Lastly, select Browse to Upload:
5. This will open the Filepicker Console as a popover window. Either click “Select Files to Upload” to choose images from your desktop, or simply drag and drop images into the console.
For these images:
- Use high-resolution photos (at least 640x480) when gathering training material for the Face Library.
- Please only use .jpg’s or .png’s.
- Use images that only have one face in the frame. Do not upload images with multiples faces.
- Ensure that both eyes are visible in the faces in selected images. Do not use pure profile shots.
6. Once added to the Filepicker, thumbnails will appear. Click on the Edit Button next to each image to initiate cropping:
7. This will open the Edit Image Console. Click on the Crop tool at the top-left to begin cropping:
8. Use the Cropping Tool to create a square crop that is close around the face. As a best practice, we recommend centering the crop around the face as much as possible, and leaving a small space around the found. Ensure that no part of the face is cut off by the crop:
9. Once satisfied with your crop, click Save at the bottom right-hand corner of the page. Then click Done. This will return you to the original Filepicker Console. Repeat process with all added images. Once all images are cropped, click Upload at the bottom-right corner of the Filepicker:
10. This will return you to the original Add New Person page. Review all thumbnails to ensure that no portions of the face are cut off by the frame. Once satisfied, click Save:
11. Once done adding new Entities to the Library, select Publish & Train. Note that you should not publish until all new Entities are added:
12. After the Library has been published, the Entity will display as Published. The Publish & Train module will also display Published:
13. This indicates that the training process has begun. You will know when training is complete when a Last Trained Date appears above Published at the top. At this point, the new faces will start being recognized by the Face Engine:
- If you want to recognize different versions of a person, train the face as two separate Entities. For example, if a person sometimes wears glasses and other times does not, you may want to train two different entities for that person (one with glasses and one without). Another instance where you may want to do this is if you’re looking to recognize a younger and older version of a person.
- Crop your images so that they are as close to the face as possible, without cutting of any portion of the face.
- Before publishing your Library, double check your thumbnails to ensure that no part of the face has been cut off by the frame.
- Note that the amount of time that it takes to train the Library will depend on the number of Entities already existing within the Library, as well as the number of Entities that were added. The Library will start recognizing new faces as soon as it is fully trained and a Last Trained Date appears.
- Manage your Entity images over time to maintain engine integrity and account for any changes in a person’s face that may happen over time.