Erreur de détection d'objets utilisant Deep Learning : la méthode .vectorize() de Python Raster Function n'a rien retourné

NexSpace

Une erreur apparaît lorsque j'utilise "Detect Objects Using Deep Learning" dans ArcGIS Pro, le modèle est usa_building_footprints.dlpk qui a été entraîné par ArcGIS Pro 2.7.3.<\/SPAN><\/P><\/span><\/P>Comportement attendu<\/STRONG>
Comment faire fonctionner le modèle.<\/P>Plateforme (veuillez compléter les informations suivantes) :<\/STRONG><\/P>OS : Win10<\/LI>Version d'ArcGIS Pro : 2.7.3<\/SPAN><\/LI>GPU : NVIDIA Quadro P620, 4GB GPU Memory<\/SPAN><\/LI>Taille de l'image test : 936MB<\/SPAN><\/LI><\/UL><\/span><\/P> <\/P> <\/P>

Tim_McGinnes

It is the error directly above that is causing the problem: RuntimeError: CUDA out of memory.

This means that all the dedicated memory on your graphics card is being used up while running the Detect Objects tool. The Batch Size parameter controls how many image segments are sent to the graphics card at once for detection. You will need to reduce the Batch Size parameter in the tool until it runs without error. I suggest setting it to 1 or 2 to begin with, and you can increase it later if required. See attached screenshot to see where the Batch Size parameter can be found.

batch_size.jpg