PeopleID automatically tags specific people in photos based on how similar the faces are to the reference headshots you provide. Just like IPTC keywords, PeopleID tags can be used to find images using search tools in the Library and on the Portal. Please contact your Customer Success Manager to add this feature to your subscription.
- Overview
- Adding people
- Editing people
- Adding new people automatically
- Adjusting the confidence
- Enabling Tagging
- PeopleID for Video (AI Video Tagging)
-
FAQ
Quality matters
Using good quality headshots for the people you want to identify is extremely important. Good quality headshots produce the most accurate results whereas poor quality headshots will result in no tags or people being tagged incorrectly.
High quality headshots are:
- Recent (within 2 years)
- Forward facing with no sunglasses, masks, hats or anything else that covers the face
- Do not contain other faces (including faces on clothing or in the background)
- Are well lit and clearly visible with no obstruction, blur, or pixelation - this includes having appropriate exposure and contrast in the photo
- Are large enough (at least 500px on the short side) - the face should fill almost all of the frame
- Are right side up (not rotated or upside down)
- JPG files (RAW and HEIC are not currently supported)
-
5 MB or less in size
Adding people
If you want PeopleID to try to identify specific people in the photos you upload, you’ll need to add each person to your list of known people/faces, following the instructions below:
Old Library |
Lumen Library |
|
|
Add a batch of people from a gallery of headshots:
To add a batch of people from photos already in your account, you’ll first need a gallery of headshots:
Create a Gallery of headshots:
- Open your Library
- Select the location where you want your new gallery to be created
- Click New Gallery
- Give your Gallery a name and click Submit
-
Upload a quality headshot of each person into the Gallery
Pro tip: If the person’s name is written into the IPTC Persons Shown field, it will be added automatically! This feature only works when using the batch action "Index Gallery" option in the steps below, not for adding individuals.
Note: For best results the gallery should contain only individual headshots (no group photos or other files).
Once the Gallery is ready, follow the instructions below:
Old Library |
Lumen Library |
|
|
Add a batch of names from a CSV:
In the event that you have a list of names of people you want tagged but don’t have all of their photos yet, you can add a batch of names using a CSV:
Create your CSV file:
- Create a new spreadsheet in Excel, Numbers, or Sheets
- Create a Name column
- Optionally, create a Metadata column to hold other information about your people such as their department or job title
- Add the name of each person to its own row in the document
- Save your document as a CSV file.
Upload your CSV file:
Old Library |
Lumen Library |
|
|
Editing people
Editing scanned face/headshots in PeopleID
Old Library |
New Library |
|
3. The name/metadata for the PeopleID record you selected will be editable, and if the user has no image associated, the option to upload an image will display. |
Rename a person
To rename a person or add a name to someone who is unidentified:
Old Library |
New Library |
|
|
Merging people
In the event that the same person shows up more than once in your list of people, you can merge them:
Old Library |
New Library |
|
|
Removing people
In the event that you don't want someone to be tagged any more, you can remove them from PeopleID. Removing someone permanently deletes that person from the system and removes all associated tags (this is a destructive action that cannot be undone):
- Open your list of People
- Find the person who you want to remove
- Click the three dots to open the Extra Options.
- Select Remove Person
-
Click Delete
Add new people automatically
PeopleID has the ability to learn new faces automatically as you upload images. With this setting turned on, when a photo is uploaded to your account, PeopleID will attempt to learn new faces, and add them to your list of people as “unidentified.”
Updating an “unidentified” person with their name will make all images they were tagged in searchable retroactive to the time they were added to your list of people. This can be extremely useful if you’re not sure who needs to be tagged in the future.
Old Library |
New Library |
|
|
Adjusting the confidence
With PeopleID, you have the ability to control how closely an image needs to match someone's face for their name to be applied as a tag.
Old Library |
New Library |
|
|
- Higher confidence settings (e.g. 95) will produce fewer, more accurate tags
- Lower confidence settings (e.g. 50) will tag people more often with potentially less accuracy, especially if several people in your list have mathematically-similar faces.
Note: This setting should only be adjusted if you’re not happy with the default behavior of PeopleID.
Enabling Automated Tagging
You'll be able to enable automated tagging in two ways: Library wide, or applying tags at the Gallery level.
To ensure that all new images are tagged upon upload, you can choose to apply automated tags to all new uploads or apply tags on the gallery level.
Old Library |
New Library |
|
Under Admin > Metadata Settings, you will see the option to "Apply automated tags to all new uploads" under Automated Tag Settings, which you can check off to enable. Click Save to make sure it is enabled.
|
Under Library > Settings > Metadata, you will see the option to "Apply automated tags to all new uploads" which you can toggle off/on. Click Save Changes when you're finished. |
- If you would rather apply PeopleID at the Gallery level, read how to enable Gallery-level tagging here.
To retroactively apply automated tags to your account for files uploaded prior to the date automated tagging was enabled, please reach out to your Customer Success Manager.
PeopleID for Video (AI Video Tagging)
Our advanced AI recognition technology extends its capabilities to video content as well. Leveraging the same comprehensive settings used for image recognition, PeopleID can effectively identify and tag familiar individuals appearing in your video assets.
It's important to note that the processing of videos differs from that of photos, which affects the billing structure for video tagging. While AI tagging for photographs incurs a charge of one credit per image—regardless of the number of tags applied in said image—video tagging is billed based on the duration of processing time, at a rate of per minute.
All PhotoShelter for Brands subscriptions include video minutes. Library staff can see the available AI Video minutes in your library alongside your available AI photo tags on the right-hand "Info" panel of your Library (New Library UI). If you require more video minutes or photo tags (or both), extra minutes or image tags can be purchased - to do so, please reach out to your Customer Success Manager. This structured approach ensures that you have sufficient resources to maximize the benefits of our AI technology in both photo and video formats.
FAQ
| How many people does PeopleID support? | PeopleID can support thousands of people. However, research has shown that artificial intelligence will confuse people once the list approaches 5,000 individuals because some people will have similar faces in a group that large. |
| Why won't PeopleID add tags when someone is wearing sunglasses or a face mask? | Approximately 50% of the data artificial intelligence uses to match faces comes from the eyes. When the eyes (or other large portion) of the face are covered, AI will have trouble making a confident match even in situations where a human being could easily identify someone with sunglasses. |
| Do I need to keep the photos that PeopleID has used to identify faces? | No, once PeopleID has been trained, it retains the face data even if the original asset used to train PeopleID is no longer present in your Library. If you wish to remove a person from PeopleID, please refer to: Editing people |