Give any robot eyes with the iPhone you already have.
Robot Eyes turns an iPhone Pro into a robot's camera and a dataset recorder at the same time, from one camera session,
so the live stream and the saved recording share one clock and one frame counter.
RECORD
• Color video (HEVC), LiDAR depth and confidence maps, accelerometer and gyroscope, and the camera's position and
orientation from ARKit, with per-frame camera intrinsics.
• The open Stray Scanner dataset layout, readable by existing tools, plus a session.json that summarizes rates,
thermal state, battery and completeness.
• Crash-safe fragmented video, a clear Complete / Incomplete state for every session, Keep pins, bulk Select ▸ Delete,
and Share with real progress and a Cancel that works.
STREAM
• Live camera over the USB cable in the Mobcam wire format (the one Moblin uses), 15 fps by default, adapting to heat.
• A once-a-second status message: frame rates, thermal, battery, free storage, camera tilt, and optionally the camera's
position, heading and speed at 5 to 15 Hz.
• A small HTTP file server for the measurement files, so a computer can grade a run seconds after it ends.
• Your computer may start and stop recordings, if you allow it.
• A ROS 2 camera adapter (Humble and Jazzy) publishes Image and CameraInfo, with checkerboard calibration support.
• Wi-Fi is off by default. Turn it on and receivers must present the access token shown in Settings.
MADE FOR THE BENCH
• Every screen carries a one-word label, so "be on the Record view" is a checkable instruction.
• A heads-up display in words, not icons: what the stream is doing, what the disk is doing, what to do next.
• Presets for a general recorder and for the Robo Bump Truck, the RC truck this app was built to drive.
PRIVACY
• No account, no analytics, no ads, no third-party code. Nothing leaves the phone unless you connect a receiver or
tap Share. Support: rooneytech.pages.dev/robot-eyes
REQUIRES
• An iPhone with a LiDAR scanner (iPhone 12 Pro and later Pro models) and iOS 18.6 or newer.
Robot Eyes is built on the open-source Stray Scanner recorder (MIT). Source and the ROS adapter:
github.com/12mv2/robo-bump-truck-ros
