|Title: PhD position - 2D complex urban area labelling||Posted: March 31, 2016|
|Location: Paris, France|
Description: We aim to label complex urban areas. Using Very High Resolution (VHR) optical images, we
focus on the most salient topographic objects, i.e., roads and buildings. In areas where a large
amount of reference data is available, it is possible to compute deep-learning based models so as
to recognize such objects. Then, these models will be propagated to areas where only few labeled
instances exist. Finally, the classification obtained with our tailored architectures will be refined
with a structured reasoning that includes detected geometrical patterns and a priori knowledge
on the interaction between main urban elements.
Application Instructions: More details here.