Presentation of our publication at QCAV 2025
Real-time object detection for sustainable edge deployment in waste sorting sites
In June 2025, Wasoria had the opportunity to present our work at the international QCAV 2025 (Quality Control by Artificial Vision) conference, a leading event dedicated to computer vision applied to quality control and industrial environments. This participation marks an important milestone in our research on embedded artificial intelligence for waste sorting applications.
Our paper, entitled “A review of real-time object detection for sustainable edge deployment in waste sorting sites,” provides a comprehensive state of the art of real-time object detection methods, with a particular focus on their deployment on embedded systems directly on site.
Why is real-time on-site detection essential?
Modern waste sorting facilities face several challenges: increasing waste volumes, stricter recycling requirements, and the need for reliable automation. Computer vision, and in particular deep-learning-based object detection, offers promising solutions to automatically identify and classify materials.
However, real industrial environments impose strong constraints:
- limited computational resources,
- real-time operation requirements,
- controlled energy consumption,
- robustness in visually complex environments.
Deployment on embedded devices (“edge devices”) makes it possible to process data directly near the sensors, without relying on heavy cloud infrastructure. This approach improves responsiveness, reduces latency, and lowers both energy consumption and infrastructure costs.
A state of the art to guide technological choices
In this article, we analyze the main modern object detection architectures, including convolutional neural network–based approaches and models optimized for real-time performance. We compare their performance in terms of accuracy, inference speed, and energy efficiency—key criteria for sustainable industrial deployment.
This work aims to provide a clear overview of the trade-offs between performance and hardware constraints, helping researchers and industry professionals choose the most suitable solutions for waste sorting environments.
A collaboration between research and industry
This publication is the result of a collaboration between the ImViA laboratory and the company WASORIA, combining academic expertise with knowledge of real industrial constraints. This partnership helps direct research toward practical solutions that can be deployed directly in the field.
The presentation at QCAV 2025 provided an opportunity to share these results with the international scientific community and to exchange with other stakeholders working on similar challenges.
Towards smarter and more sustainable sorting systems
This work is part of a broader effort to make sorting systems more efficient, autonomous, and sustainable through embedded artificial intelligence. Improving waste detection and classification is a key step toward optimizing recycling processes and reducing environmental impact.
We are proud to contribute to this effort and thank all partners who made this publication possible.
