The Apple CoreML can be divided into two implementations:
- iOS Crowd Counting Application
- macOS App (Ported from Prediction.playground because playgrounds are too fickle)
iOS Crowd Counting
Ability to drag and drop an image to see the results of different prediction strategies.
- Use Xcode 10+ for Swift 4.2, and macOS Mojave for Core ML 2 and Create ML.
- Install carthage.
carthage update --platform ios,macosor
- Build CrowdCount and CrowdCountMac. For Prediction.playground, explicitly build the libraries CrowdCountApi and CrowdCountApiMac.
- For iOS app, use Google Firebase to create a standard bare bones
GoogleService-Info.plistor just delete the file
GoogleService-Info.plistfrom Xcode. Used for image upload on request.
- Release v0.2
- Allow users to select correct classification to help model training
- Ability to select a photo from the library and segue to show page
- Have previous prediction preview and count “roll” to the left to imply that it’s working over time (animation).