Cross-platform development frameworks for mobile on-device machine learning applications
Jäntti, Larri (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025120131267
https://urn.fi/URN:NBN:fi:amk-2025120131267
Tiivistelmä
The business value of mobile implementations in the field of AI has risen over the past years. The adoption of native or cross-platform development workflows can be intimidating without prior experience, but opening a possibility for future development aspects in mobile environments should not be disregarded. While more developers become familiar with ML workflows, the models often end up being utilised through a cloud or self-hosted server APIs. Local on-device applications are still important for security and privacy focused customers. Features, possibilities and restrictions of native mobile development are well known, but the recent advances in cross-platform development options have remained somewhat unexplored.
The goal of the study was to find and recommend suitable cross-development framework(s) and target device range to focus possible future on-device ML mobile development on. For testing, similar demo applications were built with three different cross-platform development frameworks: BeeWare, Flutter and React Native. The tested ML implementation was Finnish transcription from speech to text with OpenAI’s Whisper ASR. The development frameworks were evaluated based on set criteria ranging from supported deployment platforms to availability of ML libraries and subjective ease of development. After completion, the demo apps were tested for performance. The resulting recommendations, while written with specific software development teams in mind, can be generalised to other teams or companies as well.
In the end, a promising proof-of-concept was achieved using Flutter. A working prototype was developed with React Native as well, although the same performance level could not be reached. As a result, both Flutter and React Native were recommended with a nod towards Flutter.
The goal of the study was to find and recommend suitable cross-development framework(s) and target device range to focus possible future on-device ML mobile development on. For testing, similar demo applications were built with three different cross-platform development frameworks: BeeWare, Flutter and React Native. The tested ML implementation was Finnish transcription from speech to text with OpenAI’s Whisper ASR. The development frameworks were evaluated based on set criteria ranging from supported deployment platforms to availability of ML libraries and subjective ease of development. After completion, the demo apps were tested for performance. The resulting recommendations, while written with specific software development teams in mind, can be generalised to other teams or companies as well.
In the end, a promising proof-of-concept was achieved using Flutter. A working prototype was developed with React Native as well, although the same performance level could not be reached. As a result, both Flutter and React Native were recommended with a nod towards Flutter.