Код: Выделить всё
• If I start a session right after cooling the device (e.g. refrigerated), fx/fy are noticeably higher (~3132 / 3130).
• If I start a session when the device is already warm, fx/fy are lower (~3105).
• Once the session is running, intrinsics remain consistent and do not drift over time.
• The issue only appears between sessions, where the starting baseline shifts with thermal state.
< /code>
воздействие: < /p>
• Using per-frame intrinsics works fine inside a single session.
• But across sessions, the change in baseline intrinsics introduces inconsistencies (e.g. scale mismatch in reprojection error or marker distances).
< /code>
Вопросы: < /p>
1. Is this expected behavior — that iPad intrinsics are recalculated at session initialization based on the device’s thermal/optical state?
2. Is there any supported way to “lock” or reuse the same intrinsics across multiple sessions, regardless of temperature?
3. If not, what is considered best practice?
• Recalibrate per session?
• Always trust the dynamic intrinsics?
• Or build a correction model that accounts for temperature-dependent shifts?
Подробнее здесь: https://stackoverflow.com/questions/797 ... fx-fy-base
Мобильная версия