Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : IRIMAS, Université Haute-Alsace
Durée : 6 mois
Contact : maxime.devanne@uha.fr
Date limite de publication : 2024-12-23

Contexte :
Nowadays, zoo enclosures are becoming closer to natural biotopes of wildlife animals. This implies large enclosures with biological elements such as plants and trees, and landscape elements such as rocks, hills and so on. If these new ways of designing enclosures are really improving the wellness of the hosted animals, however these ones can become hardly visible. This implies two problems :
– Frustration of visitors who want to see animals
– Difficulties for the zookeeper staff to observe the animal
Particularly, this last issue can cause a) difficulties to observe an abnormal behavior of an animal, which can delay veterinary heals if necessary and b) accident if the zookeeper has to enter into an enclosure without a clear view of the animal. To cope those problems, cameras can be installed around or inside the enclosures to monitor the animals in real-time. Particularly, thermal cameras have been proved to be very efficient in enclosures with large number of plants or even during night-time. The goal of this internship is to use multi-camera setup and data fusion to detect animals using deep learning techniques such as CNNs or YOLO.

Sujet :
The intern will have to first review the existing literature based on articles and surveys about zoo animal monitoring. Then, the goal is to select and purchase cameras (RGB, thermal, other modalities) according to the state-of-the-art, and to settle them with the help of the staff of the Mulhouse Zoo. In parallel, finding in the literature neural networks such as YOLO able to create a bounding-box prediction of the position of the animal in an image. The training of the neural network can be done using databases such as DeepFaune. Finally, data fusion can be explored to enhance the performance of the neural networks by coupling RGB and thermal predictions. GPU-based architectures will be used with Python programming.

Profil du candidat :
Final-year student in Master 2 / Engineering school (BAC+5), with an Artificial
Intelligence / Computer Vision background. Good programming skills are expected (C, C++, Python). A
first experience with camera acquisition, particularly thermal images, is good.

Formation et compétences requises :
Final-year student in Master 2 / Engineering school (BAC+5), with an Artificial
Intelligence / Computer Vision background. Good programming skills are expected (C, C++, Python). A
first experience with camera acquisition, particularly thermal images, is good.

Adresse d’emploi :
Université Haute-Alsace
12 rue des Frères Lumière
68093 Mulhouse

Document attaché : 202410230749_Master_internship_zooAI_2025.pdf