This project focuses on developing a new recycling system capable of processing any type of waste by implementing advanced Artificial Intelligence techniques. Specifically, it combines emerging Self-Supervised Learning (SSL) methods with established Deep Learning approaches for material classification. Through SSL, the system enables neural networks to independently learn and extract distinctive features, improving their ability to differentiate between various material types.

To achieve this, it is essential to build a large and comprehensive dataset that includes all recyclable materials. This process involves first capturing images of each material and then annotating them. Not only is this task highly labor-intensive, but it also requires specialised knowledge, training, and experience to ensure accurate material classification.

In summary, this project develops a new AI-driven system capable of recognising different materials and subsequently enabling their separation, facilitating an efficient and automated recycling process.

Research developed within the framework of the project «WASTESESAL: Self-learning–based vision system for the recycling of different materials», Project ACE054/22/000002. With the support of ACCIÓ.