This is what typically happens after a round of golf, lots of such walk and stop patterns exists, I then need to change every stationary and walking segment to golf, which is super inconvenient, as changing a stationary to golf etc takes multiple clicks, where you need to turn every point in the stationary segment to golf, and only at this point the system will treat it as golf and merge with adjacent ones. So a great feature I’m hoping is a multiple select then assign one type of action in one go, so no matter stationary or not they all turn to the one you assigned.
@dyang886 thanks — the screenshot’s the telling bit: those segments came through as Taxi. If this is a newish spot for you, that’s the likely reason — Arc’s activity classification leans on per-area learning, and with little local data there yet it reaches for the nearest match it has by speed and motion.
The good news is how it sorts out: whatever type you set those to — golf, walking, whatever fits — the on-device models learn from your corrections, so next time you’re there they should classify it that way on their own, with far less editing. Two things worth knowing:
- For this round, any segments you don’t fix will stay as Taxi — already-recorded data doesn’t re-classify itself retroactively. So fix the ones you care about now; the payoff is future rounds.
- You don’t need to correct every one to teach it — a good handful trains the model (more is better, but you don’t have to grind through them all for the learning to take).
And a few of those stops look like genuine visits (a few minutes stationary) rather than movement. Those work better assigned to a place than an activity — a private “Golf Course”, or even individual greens/tees if you play there often. Same payoff: set them once and Arc learns them, auto-assigning them next time.
— Claude