Update README: timestamps and calibration

README.md

@@ -121,7 +121,11 @@ VISTA/
 │           │   ├── magnetometer.csv
 │           │   ├── barometer.csv
 │           │   ├── rgb_timestamps.csv
-│           │   └── time_domain_mapping.csv
+│           │   ├── slam_left_timestamps.csv
+│           │   ├── slam_right_timestamps.csv
+│           │   ├── eye_timestamps.csv
+│           │   ├── time_domain_mapping.csv
+│           │   └── calibration.json
 │           └── mps/
 │               ├── eye_gaze/
 │               │   └── general_eye_gaze.csv
@@ -141,6 +145,14 @@ VISTA/
 
 This is a representative layout, not a complete listing of supplementary metadata files. The number of `qN.wav` and `gN.wav` files varies with the annotations. 
 
+## Sensor streams, timestamps, and calibration
+
+The `vrs/` folder of each clip holds the streams extracted from the original Project Aria recording.
+
+- **Frame timestamps** (`rgb_timestamps.csv`, `slam_left_timestamps.csv`, `slam_right_timestamps.csv`, `eye_timestamps.csv`): each file has one row per frame of the matching video (`<clip_id>.mp4`, `slam_left.mp4`, `slam_right.mp4`, `eye.mp4`), with columns `frame_index,capture_timestamp_ns,timestamp_source`. `frame_index` is the 0-based frame index in the video. `capture_timestamp_ns` uses the device clock, the same clock as the IMU, magnetometer, and barometer CSVs and the MPS outputs. `timestamp_source` is `vrs` when the timestamp comes from the recording. It is `extrapolated_fps` for the few RGB frames (257 frames across 113 clips) that have no recorded timestamp; these frames were filled in the video, and their timestamps are extrapolated from the frame rate.
+- **Clock conversion** (`time_domain_mapping.csv`): pairs of device-clock and Unix timestamps (`monotonic_timestamp_ns,real_timestamp_ns`) for converting to wall-clock time.
+- **Factory calibration** (`calibration.json`): the device's factory calibration in the Project Aria JSON format. Load it with `projectaria_tools.core.calibration.device_calibration_from_json_string`. It contains intrinsics and extrinsics (`T_Device_Camera`) for the RGB, SLAM, and eye-tracking cameras, and calibrations for the IMUs, magnetometer, barometer, and microphones. The device frame is defined by `camera-slam-left`. Recordings come from four devices, whose serial numbers are replaced by `device_1` to `device_4`; clips recorded with the same device share an identical file. Camera calibrations are expressed in the raw sensor orientation. The RGB and SLAM videos are rotated 90° clockwise relative to the raw sensor images, so rotate the camera calibration accordingly (for example with `projectaria_tools.core.calibration.rotate_camera_calib_cw90deg`) before projecting onto these videos. `eye.mp4` is not rotated.
+
 ## Annotation schema
 
 Each CSV row represents one clip, identified by `clip_id`.