Optimize Workflow for Object Detection
Hi guys,
Okay, this is gonna be quiet a long story. TL:DR, how to optimize object detection using deep learning in a large area (using Esri World Imagery).
My goal is to detect palm oil boundary (image segmentation) in half of Indonesia (about 819,200 km2) using Esri World Imagery service as the raster. Since I know it will take a long time, I've tried to optimize the workflow by doing image classification for every 640x640 m grid. The grid will be classified into palm area or other, and then dissolved the grid_palm. The plan is to run the image segmentation only on area classified as palm area.
But I'm afraid that actually the process only take the extend (xmin, xmax, ymin, ymax) of the whole grid (which will make the image classification useless). Is it actually works like that? Or do you have any recommendation how to optimize the process?
Regards,
Naufal
