Tip #46Kinesense LE

How to mark faces, batch-enrol people and identify a face

Mark faces the server missed, enrol many people at once from a JSON file, and identify a face from an external, clarified or tagged image.

Patricia McCabeSoftware Engineer

To conclude our series on face recognition, we’re going to tackle batch enrolment, manually marking a face and where you can identify a face in Kinesense. Previously we covered importingand reviewing.

Manually Mark Face

If you find a face while reviewing a video that the facial recognition server hasn’t picked up you can manually make it using the Manually Mark Face button on the left hand side of the review pane.

An arrow pointing to the Manually Mark Face button in the review toolbar.

Selecting this will allow you to draw a box around the face you want.

A face being boxed by hand, then shown with an Unknown label.

A box will be drawn around the face with the header Unknown and a new face sprite will appear on the timeline. If you know who this suspect is you can add matches for it like wediscussed last week.

Batch Enrollment

Enrolling one person at a time into the facial recognition database is fine if you only have one or two people you need to enrol but is an utter pain if you have tens or hundreds or even thousands of faces to do. Instead you can create a JSON file with the information about each face and links to images, and use the batch enrolment process to add them all to the database at once.

The Options menu with Face Recognition expanded to show Batch Enrollment.

The Batch Enrollment option can be found in the Options menu under Face Recognition. This will bring up the Batch Enrollment window.

The Batch Enrollment window with the input JSON file, output folder and Allow Duplicate Names options.

The first thing you’ll need to add is your JSON file. Each face will need details about it, like name, gender, date of birth etc, and the paths of the images you wish to use.

A sample JSON entry for a person, with name, gender, date of birth, nationality, ethnicity and image file paths.

The format of this file needs to be correct so we can provide you with a sample file that you can edit with the correct details.

The next thing you’ll need to provide is an output folder. While the faces are enrolling, the database will create logs and store them in your output folder.

You can also select whether or not you want duplicated names your in database. This will mean that any duplicates wont be enrolled.

Once you’ve filled in everything click Begin Enroll and the faces will begin enrolling into your database.

Identify Face

Identify Face button

The Identify Face button can be used to identify a face in an external image, a clarified image or an image of someone you’ve tags. For an external image, use the Identify Face button in the Review tab.

The Identify Face window with a selected image and a list of possible matches with confidence scores.

Click Select Image To Identify and browse to your image. A list of results will be displayed on the right hand side.

The Identify Face window searching the database with a clarified image.

When you’ve clarified an image you can select the Identify Face button in the Clarification tab and it’ll use that clarified image to search the database.

A tag object image with an arrow pointing to the Identify Face button beneath it.

That function can also be applied to images you have attached to an object in tags.

That should do it for facial recognition. So get out there are start recognising some faces!

Format support

3000+ video formats

Anything playable in Windows Media Player, VLC or FFmpeg is supported, without the manufacturer's original player. Have a format that isn't listed? Our engineers have 17 years experience of dealing with difficult formats.