A Software Service for the Garbage Type Recognition Based on the Mobile Computing Devices With Graphical Data Input

1
Національний університет «Львівська політехніка», кафедра електронних обчислювальних машин
2
Lviv Polytechnic National University
3
Lviv Polytechnic National University

The article describes problems of determining the type and automatic sorting of household waste using mobile computing devices. All of the required hardware and partially software, required for implementation of this service, are already present in modern smartphones. iOS and Apple products were selected as the base for the service, due to such advantages over competitors: dual or triple depth camera (TDCS), powerful GPU, Neural Engine coprocessor, high autonomy (2750mAh battery size), sensors that allow for user positioning and navigation in space (GPS, Glonass, Gyroscope) and most important feature is possibility of cross-platform designing, suitable for iOS and macOS (Project Catalina). The recognition process consists of several phases, including capturing of graphic image and detecting the object shape, shape analysis, computing the results, and saving new associations to the database. The analysis itself is implemented using a neural network that is able to learn during its operation. Initially, the algorithm is driven by the selection of photographs with a certain type for the base set of associations, each subsequent scan improves accuracy. Cross-platforming plays a very important role — it allows us to develop a single software service that is initially run on a macOS-based computer for faster learning and then can be easily used on an iOS mobile device. After identifying a particular type of garbage, the route to the nearest recycling point of such type of garbage will be proposed for user or user’s clarification will be requested. User can also manually browse categories and related items, manually search by name of item, and view locations for sorting and recycling in appropriate city. When a completely unknown object arrives, it is possible to refine the information in order to help further learning of the network.

  1. Bachynskyy R.V. (2006). Selection of the image compression processor structure for typical algorithms Lviv Polytechnic National University Journal. — P. 3–8.
  2. Morozov Y.V. (2018). Basic principles of PLO Apple Inc.
  3. https://learn.ztu.edu.ua/mod/page/view.php?id=7255
  4. Apple Developer Documentation (2018). Apple Inc. — https://developer.apple.com/documentation.
  5. Map Kit (2018). Apple Developer Documentation Apple Inc. — Access mode: https://developer.apple.com/reference/mapkit.
  6. Encoding and Decoding Custom Types | Apple Developer Documentation (2018). Apple Inc. — Access mode: https://developer.apple.com/documentation/foundation/archives_and_serial...
  7. Generics in Swift 4/Candost Dağdeviren (2017). Access mode: https://theswiftpost.co/generics-swift-4/
  8. Overview Geocoding API Google Developers (2018). Access mode:https://developers.google.com/maps/documentation/geoc oding/intro?hl=en
  9. Core ML Apple Developer Documentation (2018) Access mode:https://developer.apple.com/documentation/coreml.
  10. A deep dive into Grand Central Dispatch in Swift / Jonh Sundell — 2017 Access mode: https://www.swiftbysundell.com/posts/a-deep-dive-into-grand- central-dispatch-in-swift.