FPFelipe Arruda Pontes

Felipe Arruda Pontes

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Felipe Arruda Pontes has a total of 4 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Real-Time Context-Aware Early Filtering for High-Definition Video Analytics on Commodity Edge Devices Using GenAI for Data Augmentation

    by Felipe Arruda Pontes, Michael Schukat, Edward Curry

    This work proposes a fast and accurate early filtering pipeline for video analytics in commodity Edge devices for Smart-Cities applications. This pipeline can run in real-time even on a small and GPU-less device such as a Raspberry Pi, while maintaining a good accuracy for video analytics queries. In addition to a novel Edge optimised pre-processing method, the pipeline uses a context-aware binary model, which is fine-tuned using semi-automatic synthetic data augmentation, Generative AI, and Cut-and-Paste techniques to contextualise the model to the input camera background and the Objects of Interest (e.g., car or person) from a user’s video analytics query, in a fast process that requires only 10 seconds of original footage for training. This makes it the first Edge filtering with specialised models with a viable online training solution. Compared to a baseline state-of-art Nano-YoloV5 model, the proposed early filtering pipeline in its high speed profile shows an 48.8x increase in speed and is the first of its kind that is able to run on physical hardware (i.e., non-simulated) commodity Edge devices at more than 80 FPS in HD (1920x1080) resolution, with a small accuracy loss of 5% compared to the baseline. On the high accuracy setting the pipeline still runs at more than 41 FPS (26.9x faster than Nano-Yolo) and shows an increase of 2.5% in accuracy.


  • A Comparative Study of Open-Source Fuzzy Modelling Toolkit Licenses and Features

    by Felipe Arruda Pontes, Edward Curry, Michael Schukat

    The practice of distributing software as open-source, which serves as a pillar for Open Science, can benefit future research in the field of fuzzy logic by enabling access and increasing the maintainability of free fuzzy logic frameworks alternatives to closed commercial solutions. Most open-source software uses a distribution license, with a wide range of restrictions and requirements on the distributions of these tools and their derived works. Therefore, understanding each license limitation and requirement is essential to ensure an ethical research process and avoid legal issues from breaching the open-source license. However, there is a lack of comparative research on the current state of open-source licenses for fuzzy modelling toolkits and frameworks. Previous studies are outdated or incomplete, missing some of the most popular open-source fuzzy frameworks or a detailed analysis of the licenses used. Our research analysed 20 of the most prominent open-source toolkits and categorised them based on their licenses, with qualitative comparisons of their characteristics and limitations. In our study, we note a concerning scenario where a potential license violation may impede the publication and dissemination of academic works derived from fuzzy toolkits using the GPL license. Additionally, we identify a gap in open-source fuzzy toolkits with support for Fuzzy Control Language (FCL) and Type-2 fuzzy sets. Finally, this study aims to support and guide future researchers in selecting the fuzzy toolkit that best matches their research problem without infringing the open-source licenses used, with an intuitive and up-to-date overview of the licenses and features of open-source fuzzy modelling frameworks available.


  • Fuzzy vs. Crisp in Uncertainty-Aware Service Selection: Enabling Sustainability on Multimedia Event Processing

    by Felipe Arruda Pontes, Michael Schukat, Edward Curry

    The high energy consumption projections of Cloud/Edge applications urge the development of ecologically sustainable Multimedia Event Processing (MEP) systems. In these applications, a Multi-Criteria Decision Making (MCDM) problem must be solved when selecting the best service workers alternatives for processing user queries, according to the user-specific performance criteria requirements of energy, accuracy and speed. Moreover, fuzzy logic provides a well-established method, such as the fuzzy TOPSIS, for dealing with the uncertainties arising from real-world scenarios, where ambiguities of user requirement interpretations and imprecision of measurement of the computing devices may directly impact this decision-making process. However, this fuzzy method is more complex and computationally intensive than the original (crisp) TOPSIS. Therefore, it is crucial to understand to what degree the fuzzy and crisp methods may be used interchangeably and still get the same results to avoid unnecessary complexity in sustainable MEP solutions in a real-world context. In our work, we developed a fuzzy TOPSIS ranking method for handling the uncertainties of the user criteria weights and service worker alternative ratings. Contrary to a previous study, we provide evidence that replacing the fuzzy TOPSIS method with its crisp counterpart significantly affects the ranking results when applied to a real-world scenario, with contradiction rates higher than 60% in most scenarios explored, which suggests that it is not viable to interchange these methods without consequences to the sustainability efforts of an MEP application.