- Prof. Jean Marie Dembele, Professeur Assimilé à l’Université Gaston Berger de Saint-Louis, ACE MITIC, Sénégal
- Prof. Eugène Ezin, Professeur titulaire à l’Université d’Abomey-Calavi, ACE SMIA, Bénin
- Prof. Christophe Cambier, UMI UMMISCO, IRD
- Théophile Bayet, IRD Dakar
Title: Deep learning for environmental monitoring; illegal landfills
Project type: Research Project (4 years)
This project addresses environment management with DeepLearning and Drone images; more precisely the monitoring of clandestine small landfills. Indeed, despite the collection of domestic wastes by municipal services and their deposit in official landfills, one can still find illegal landfills overall African countries. That raises challenging health and environmental problems. It is therefore urgent to identify in real-time those illegal landfills keeping them from emerging and proliferating. These monitoring, geolocation and characterization can also give a best way to handle illegal landfills at a minimal cost.
Therefore we use a deep convolutional neural networks (CNN) that extracts features from pictures or videos and provides with recognition. It is well known that with CNN, one can recognize patterns in a specific framework but in this project we want to demonstrate the possibility of indirect encoding of the weights of deep neural networks made up of thousands of connections. This will allow from deep networks making image recognition (like the one that already detects wastes) to evolve into a hyper network bringing together different patterns and being scalable with the possibility of integrating unknown patterns like precarious or unfinished constructions, anarchic or illegal land uses, crops growth, etc.
Automatic flight plans for the drone in respect to the monitoring goal (clandestine small landfills, precarious or unfinished constructions, anarchic or illegal land uses, crops growth…) is another objective of this project that involves so far partners from Senegal/Benin/France and Phd/Master Degree students.
- Research themes in digital science: Artificial Intelligence, DeepLearning, Image recognition, NeuroEvolution, Evolutionary algorithms.
- Other research themes and application areas: Environment, illegal landfills and land uses, satellite and drone images, surveillance and decision making.