Help us training Telraam v2

As we’re moving to an AI-based detection algorithm, we need video material to train the neural network. This is where you can help! If you decide to help us, great! We would need clips of 15-30 minutes from the location where your Telraam is now operating.

Ideally we need video’s in different conditions: heavy/normal traffic, rainy/windy/sunny conditions, light/dark,… Use a GoPro, a smartphone or any video-camera, whatever is most convenient for you, but take care of a few settings in order to maximize the value of your contribution. Use a normal frame rate (FPS) of 24 or 30 FPS (avoid using high FPS of 60 or above), and 720p or 1080p (FullHD) resolution. Avoid using very low or very high resolutions, such as anything below 720p or at or above 4K.

Use a normal wide angle view (for GoPro users, “Linear” is a great setting, for smart phone users the standard normal camera will be good, just avoid super wide or tele lenses) whenever possible. … It’s important that the video is taken from the location where the sensor is/would be installed, you can easily set up a smartphone or a GoPro against the window on top of a stack of books, leaning the camera against the window to achieve the same field of view that a Telraam would have (handheld recordings are not recommended).

Is this legal? Yes! You’re allowed to capture video from the public domain, if people passing by are not the main subject of the video, but simply passing coincidently. The video material will not be published and will only be used to train an AI-script and deleted afterwards. The AI-script’s intention is to detect modes of traffic, so nothing spooky like facial recognition or number plate detection, to ensure privacy.

You can send us the video through WeTransfer addressed to video@telraam.net.