Importing and Registering People
There are three main ways to register people to SAFR's Person Directory: cameras, photos, and recorded video. Imported people are registered to the Person Directory and stored in the directory specified in the User Directory setting of your Account Preferences.
- Select the connected camera you want to use by clicking on File > New to open the Desktop Client's Camera window, then select a camera from the Select Camera menu.
- Set your video feed processing mode (located in the upper right hand corner) to one of the following modes: Recognition, Import, or Learn and Monitor. Recognition is considered the default mode for set-up validation and experimentation.
- All the faces in view of the camera will initially have a grey overlay, which indicates one of two things:
- Your client or console hasn't received a response from its attempted recognition from the SAFR Server yet, or
- The face does not meet minimum image quality metric values and recognition cannot be successful. If the grey overlay persists, then the problem is the image quality. Try cleaning your camera lens or adjusting your camera placement to fix the problem.
- When the overlay turns purple, the face has sufficient image quality for recognition by SAFR but it isn't recognized by SAFR because it hasn't been registered yet. To register the face, double click the face and the Register dialog will open.
- You can choose to enter a name for the face if you want, but it's not required. Click Register to complete the registration.
- The color of the face's overlay will change to either green (if you named the face) or blue (if you didn't name the face). Both of those colors mean that SAFR recognizes the face. For more information on overlays, see Interpret Video Feed Overlays.
Another way to register faces is by using a Mobile Client installed on an iOS or Android device. For more information, see the Create an Android Registration Kiosk or Create an iOS Registration Kiosk documentation.
To import faces from picture files, do the following:
- Open either the Desktop Client (by clicking on the SAFR icon on your desktop) or the Web Console. (See Access the Web Console for information on how to do this.)
- On the Desktop Client, open the People Window by clicking on Tools > People. On the Web Console, click on the People tab.
- On the Desktop Client, click the Add face button at the top of the People Window, then select a facial image. On the Web Console click on the up arrow symbol in the upper right hand corner and select a facial image. Note: On the Windows Desktop Client, you're able to select multiple facial images at once, thus allowing you to register multiple people at once.
- Image files are usually .jpg, .jpeg, or .png files. If the file you selected has multiple faces on it, SAFR will attempt to import all the faces in the image.
- On Windows, you're able to select multiple image files at the same time to import all of them at one time. We recommend that you select no more than 20 images at a time. If you select more than ~20 images at a time, you're likely to experience a degradation of system performance.
- When you import facial images, you may be prompted to resolve any duplicate and/or low-quality image conflicts that may have arisen.
Windows only.
To resolve conflicts resulting from people that already exist in the Person Directory, do the following:
- Click on the Fix conflicts link in the notification bar.
- Duplicates are displayed side by side in the Report Dialog.
- You will be given 4 options:
- Create New Person with Imported Face: A new identity will be created, and the imported face will be used for the identity's reference image.
- Replace Existing Face with Imported Face: The imported face will replace the already-existing face as the reference image for that identity.
- Add to Existing Person: The imported face will be added as an alternate face image for the identity.
- Skip Import: The new face isn't imported; it will be discarded.
- Repeat the previous step for all duplicates.
Note: At the top of the Report Dialog, there is an option to accept the default recommendations. SAFR defaults to accept whichever of the duplicate images is higher quality based on internal image quality metrics.
Windows only.
To resolve conflicts resulting from low-image quality images, do the following:
- Click on the Fix conflicts link in the notification bar.
- Low-quality images are displayed in the Report Dialog. Note that a warning symbol appears next to the image quality metric(s) that are problematic. See Image Quality Metrics Guidance for information and guidance on the metrics.
- Click on Import anyway to import the low-quality image, or click Skip Import to not import the image.
- Repeat the previous step for all low-quality images.
Note: Clicking on X in the Notification bar or at the top of the Report Dialog cancels the import operation for all remaining low quality image conflicts.
You can automate the process of registering people by using the safr_people_import_cli.py script, which can be downloaded here.
To run this script, you must have Python installed. You must also have the following Python dependencies installed:
- requests
- pandas
- xlrd
- openpyxl
The easiest way to do this is to download requirements.txt and run the following command with administrative privileges, either by opening an elevated command prompt (in Windows) or by prefacing the command with sudo
(in Linux):
python -m pip install -r requirements.txt
To get instructions on how to run the script, run the following command in the same directory to which you downloaded the safr_people_import_cli.py script:
python safr_people_import_cli.py -h
You have the option of automatically assigning the metadata for the images by passing the script a CSV or XLS file containing the images' metadata elements separated by commas, one image per line. On each line the metadata elements must appear in this order:
imageFilename,firstName,lastName,name,age,dateOfBirth,gender,externalId,expDate,validationPhone,validationEmail,tags,homeLocation,personType,moniker,company,idClass,registrationSource,registrationSite
Any elements that you don't want to assign should simply be omitted from the element list. For example, if you only had the first name, last name, and idClass for an image, its line in the CSV file might appear as:
AnImage01.jpg,Bob,Johnson,,,,,,,,,,,,,,Threat,,
You can open a saved video file to recognize and extract facial recognition data. To do so, do the following:
Open the Desktop Client.
Click File > Open, and then browse to any saved *.mp4, *.webm, or *.mkv file to open it.
If you're on a Windows machine and you have event reporting enabled for the currently selected video processing mode, (located on the Events Preferences tab) the dialog below will open. (If you don't meet both of these conditions, then the video will simply open.)
- Actual start time: The timestamp that the video will acquire when you press Play. (e.g. In the example above, the played video's timestamp would start at 12:38,10/28/2019) The input box starts ‘live' and keeps up-to-date with the local time. When you interact with the time or set the focus, the input box stops being live.
Note: Deleting the timestamp and leaving the field blank is valid, despite the red outline that the field acquires. Of course, if you do leave the field blank, the video won't have a timestamp, as expected.
- Site: The Site label that will be applied to all events generated by the video. This field is auto-populated with your User Site preference located in the Account Preferences.
- Source: The Source label that will be applied to all events generated by the video. This field is auto-populated with the name of the video.
Set the video file's video processing mode to Recognition.
SAFR will proceed to register any unregistered faces that appear in the video.