Performance Evaluation of Cross-Platform AI-Powered Content Moderation Systems: Flutter vs Native Android
DOI:
https://doi.org/10.51137/wrp.ijmat.774Keywords:
Content Moderation, Cross-Platform Development, Flutter, Android, ONNX Runtime, Performance Evaluation, Mobile AIAbstract
The following study evaluates the performance metrics of cross-platform and native implementations of AI-driven content moderation services. Due to the rapid growth of user-generated content available on mobile platforms, real-time content moderation became necessary. To contrast the performance metrics of the cross-platform implementation of the screen monitoring system using Flutter framework and native Android application that employs the same algorithms for image analysis (ONNX Runtime) and OCR recognition (ML Kit). Both apps were tested for 15.5 minutes each and recorded the following metrics – CPU usage, RAM consumption (PSS), battery discharge, and processing time per analysis frame – across ~ 900 frames. The CPU usage rates were practically equal (Flutter – 18.91%, Android – 19.42%). Average memory consumption was estimated at 955.3 MB for Flutter (PSS) compared to 180 MB in total for Android (PSS). Discharge rate was also similar (Flutter – 192 mAh; Android – 197 mAh). Average processing time for one analysis frame was 610 ms with a range of 195–1631 ms. Despite being equally computationally efficient, Flutter has considerably more memory consumption, which is caused by its rendering engine.
References
Statista. (2026). Number of smartphone mobile network subscriptions worldwide from 2016 to 2028 [Graph]. Retrieved May 11, 2026, from https://www.statista.com/statistics/330695/number-of-smartphone-users-worldwide/
Bajer, L. (2025). Mobilní aplikace pro události pořádané v okolí.
H. Mohammed. (2024). Understanding Flutter’s Rendering Engines: Skia vs. Impeller. Medium.
Biørn-Hansen, A., Rieger, C., Grønli, T. M., Majchrzak, T. A., & Ghinea, G. (2020). An empirical investigation of performance overhead in cross-platform mobile development frameworks. Empirical Software Engineering, 25(4), 2997-3040.
Barros, L., Medeiros, F., Moraes, E., & Júnior, A. F. (2020, July). Analyzing the Performance of Apps Developed by using Cross-Platform and Native Technologies. In SEKE (pp. 186-191)
Pocatilu, P., Vetrici, M., & Despa, M. L. (2020). Performance analysis of cross-platform mobile frameworks for compute-intensive applications. IEEE.
Shreeman, D. (2025). Nitro: Zero-overhead FFI for Flutter. https://nitro.shreeman.dev/
Flutter Team. (2020). Channels are ~2x slower on Android versus iOS [Issue #53311]. GitHub. https://github.com/flutter/flutter/issues/53311
R. Gupta. (2023). Speed up inference of your hosted LLMs: Quantization. Neural Notions.
Aim Techno Labs. (2025). TensorFlow Lite vs. ONNX Runtime: Choosing an engine for your edge AI project. Technical Report.
Jhaver, S., Zhang, A. Q., Chen, Q. Z., Natarajan, N., Wang, R., & Zhang, A. X. (2023). Personalizing content moderation on social media: User perspectives on moderation choices, interface design, and labor. Proceedings of the ACM on Human-Computer Interaction, 7(CSCW2), 1-33.
Chou, H. Y., Naveed, W., Zhou, S., & Yang, X. (2026). Are Open-Weight LLMs Ready for Social Media Moderation? A Comparative Study on Bluesky. arXiv preprint arXiv:2602.05189.
Malavolta, I., Soru, S., & Salehi, M. (2021). An empirical investigation of performance overhead in cross-platform mobile development frameworks. Empirical Software Engineering, 26(4), 1–32.
Published
Issue
Section
License
Copyright (c) 2026 Ahmed Ibrahim Abd ElGhany, Mohamed Abd Elfattah, Noha Ayman, Rokiya Abd ElSatar, Kerlos Fatouh, Ahmed Abbas, Hasnaa Nageh (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
