The following classes are available globally.
Activity Type Classifiers are Machine Learning Classifiers. Use an Activity Type Classifier to determine the
PreconditionAn API key is required to make use of classifiers. See
Supported Activity Types
stationary, transport, walking, running, cycling
Base types match one-to-one with Core Motion activity types, with the exception of Core Motion’s
automotivebeing renamed to
car, train, bus, motorcycle, airplane, boat
Extended types are a subset of the base
transporttype, allowing for more specific classification when enough local data is available.
Region Specific Classifiers
LocoKit provides geographical region specific machine learning data, with each classifier containing the data for a specific region.
This allows for detecting activity types based on region specific characteristics, with much higher accuracy than iOS’s built in Core Motion types detection. It also makes it possible to detect a greater number of activity types, for example distinguishing between travel by car or train.
LocoKit’s data regions are roughly 100 kilometres by 100 kilometres squared (0.1 by 0.1 degrees), or about the size of a small town, or a single neighbourhood in a larger city.
Larger cities might encompass anywhere from four to ten or more classifier regions, thus allowing the classifers to accurately detect activity type differences within different areas of a single city.
Determining Regional Coverage
Stationary, Transport, Walking, Running
The base activity types of stationary, transport, walking, and running do not significantly differ by geographical region, thus should achieve high detection accuracy everywhere in the world, regardless of local data availability.
These types can be considered to have global coverage.
Cycling has enough regional variance in locomotive characteristics that detection accuracy can range from excellent to average depending on the availability of local model data.
If very high accuracy cycling detection is important to your application, you should check the cycling coverage map for the regions you require. However if cycling detection is not a core function of your app, then the results from even low coverage classifiers should achieve adequate accuracy, and will certainly exceed the accuracy of Core Motion’s detection.
Car, Train, Bus, Motorcycle, Airplane, Boat
While the baseSee more
transporttype can be detected anywhere in the world with high accuracy, determining the specific mode of transport requires local knowledge. If knowing the specific mode of transport is important to your application, you should check the coverage maps for your required regions.
public class ActivityTypeClassifier: MLClassifier