jakies pierdoly

README.md

@@ -109,14 +109,18 @@ Each run writes `results/gaze_exercises_<timestamp>.csv` plus a sidecar
 every recording with `--subject-id` and, when varying setup, `--session-tag`:
 
 ```powershell
-python gaze_exercises.py --subject-id you --session-tag laptop_50cm --screen-diagonal 15.6 --viewing-distance 50
-python gaze_exercises.py --subject-id you --session-tag monitor_90cm --screen-diagonal 27 --viewing-distance 90
+python gaze_exercises.py --list-monitors
+python gaze_exercises.py --subject-id you --session-tag laptop_50cm --monitor 1 --viewing-distance 50
+python gaze_exercises.py --subject-id you --session-tag monitor_90cm --monitor 2 --viewing-distance 90
 ```
 
-`--screen-diagonal` (inches) and `--viewing-distance` (cm) must match the
-physical setup for that recording — they drive the pixel-to-degree conversion
-used for both the logged error and velocity columns, so an incorrect value
-silently skews every derived feature in that session.
+`--monitor` picks which physical display to run on and auto-detects its
+diagonal; pass `--screen-diagonal` explicitly only if auto-detection reports
+`unknown size` for that display. `--viewing-distance` (cm) has no way to be
+auto-detected and must match the physical setup for that recording — it
+drives the pixel-to-degree conversion used for the logged error and velocity
+columns, so an incorrect value silently skews every derived feature in that
+session.
 
 CSV columns beyond the basic error metrics:
 
@@ -162,6 +166,14 @@ Once more subjects are recorded, re-run training; group the split by
 `subject_id` (leave-one-subject-out) instead of by trial for the real paper
 evaluation.
 
+`dispersion_px` and `accel_px_s2` in the raw CSV are in screen pixels, which
+are not comparable across recordings taken on different physical screens or
+viewing distances. Training re-derives both in visual-angle units
+(`dispersion_deg`, `accel_deg_s2`) using each session's `_meta.json`, matching
+what `adaptive_filter.py`'s `OnlineFeatureTracker` computes online — this
+matters as soon as you record on more than one screen/distance combo, which
+is exactly what varying `--monitor`/`--viewing-distance` across sessions does.
+
 `adaptive_filter.py` defines the real-time filters built on that model:
 `BaselineFilter` (the existing fixed median+EMA filter), and
 `MLAdaptiveFilter` in `mode="hard"` (discrete fixation/saccade/pursuit
@@ -177,7 +189,7 @@ reports RMS error, mean error, and on-target accuracy % for raw gaze, the
 fixed baseline filter, and both ML-adaptive modes side by side:
 
 ```powershell
-python eval_static_fixation.py --subject-id you --duration 60 --viewing-distance 70 --screen-diagonal 31.5
+python eval_static_fixation.py --subject-id you --duration 60 --viewing-distance 70 --monitor 1
 ```
 
 Press Space to start the 60-second hold. Results are saved to