-
Notifications
You must be signed in to change notification settings - Fork 425
AI Face Tracking
The AI models used for face trackers made a lot of progress and became very good in recent years. There are reports of people, who already own dedicated hardware such as TrackIR, but use Opentrack and an AI face tracker instead, because they don't have to wear any special equipment on their head and it performs such as well for them.
However, if face tracking doesn't work reliably for you, you can also try the Aruco tracker. If you want more reliability (especially in dark environments), use a setup using the point tracker and an IR clip, like the ones sold by Delanclip (UK), Grass Monkey Simulations (USA) or TrackHat (UK).
Neuralnet
- free and open-source software
- included in Opentrack, no additional installation necessary
- uses a normal webcam or PS3 Eye camera for tracking
- there are reports that it performs better than the older AITrack model
- more info: github.com/opentrack/neuralnet-tracker-traincode
AITrack
- free and open-source software
- uses a normal webcam or PS3 Eye camera for tracking
- more info and installation: github.com/AIRLegend/aitrack
Eyeware Beam
- paid and proprietary software
- uses a normal webcam for tracking
- besides head tracking, it also does eye tracking and combines both
- alternatively they also offer an iOS app
- more info and installation: beam.eyeware.tech
- see the wiki for the info about iOS app
SmoothTrack
- paid and proprietary software
- runs on iOS and Android devices and uses their native AR capabilities
- uses the device camera for tracking and sends data over the network to opentrack
- can be used with USB cable instead of WiFi to reduce latency
- more info and installation: smoothtrack.app
Tobii Eye Tracker
- can be used with Opentrack or its own proprietary software
- uses a proprietary device as camera
- there are reports it has slightly higher latency than other trackers
- besides headtracking it also provides eye tracking
- more info: gaming.tobii.co
Although most software using webcams work with pretty much any webcam, high refresh rate (60+ FPS) webcams usually perform better tracking. These can be quite expensive. Therefore it is widely recommended to use a PlayStation 3 Eye camera, which are are specifically made for tracking (they have a very high refresh rate on low resolutions) and are commonly sold on Ebay used for about 10 USD/EUR. An open-source driver for the PS3 Eye camera is released by Opentrack.
For the PS3 Eye camera the most recommended settings are 640x480@60Hz with MJPEG enabled. The field of view (FOV) depends on which zoom level the lens barrel is rotated at. On the very left side (red dot) the FOV is 56 and on the very right side (blue dot) the FOV is 75.
Apps like DroidCam can also be used to use a phone or tablet as a webcam. It is recommended to connect the phone to the PC via USB instead of using WiFi for lower latency.
Although it is recommended to tweak your settings manually to achieve the best performance, here is a list of guides and settings for orientation.
General:
- Quick-Start-Guide
- Accela Filter Wiki Page
- Guide for Head Tracking (IL2-Forum)
- Common Camera FOV Values (AITrack Wiki)
Neuralnet tracker:
- Delanclip publishes preconfigured config files in their face tracking instructions (change camera settings if you don't use the DelanCam1)
- User submitted config for the Neuralnet tracker with a Logitech C920
- Another user submitted config for the Neuralnet tracker with a Logitech C920
AITrack tracker:
SmoothTrack: