Calibration par DeepLearning

 Stage · Stage M2  · 6 mois    Bac+5 / Master   Cirad · Montpellier (France)  ~575 euros/mois

 Date de prise de poste : 1 mars 2022


DeepLearning Calibration Plant growth Voxnet


RoCoCau is a plant growth simulator that can produce 3D shape of realistic plants along time including both shoot and root compartments. A plugin named TOY was added to RoCoCau to express the functional dependency between these two compartments for biomass production and sharing depending on environment quality.

TOY has 27 parameters among which 18 cannot be measured and thus has to be estimated thanks to numerical tool. Preliminary studies have been led about the use of Voxnet convolutional network to train and predict the values of these 18 hidden parameters. Results were good enough to continue further on improving this method.

The goal of this internship is to improve the parameter calibration while focusing on network structure, input/output and distance measurement along learning process.

Internship will take place in Cirad - Montpellier - France.


Procédure : Envoyer un mail a

Date limite : 3 janvier 2022


Jean-Francois Barczi

Offre publiée le 10 novembre 2021, affichage jusqu'au 3 janvier 2022