Closed nathanielrindlaub closed 3 years ago
Also get in contact with Nanolayers developers to ask about the megadetector->mira pipeline they put together. Were they cropping the input images to the detection bounding boxes and just submitting the cropped image to MIRA? Is there any other preprocessing going on (resizing, scaling, etc.)? Did they document performance gains against a baseline?
github.com/tnc-ca-geo/mira