Improving object detection using deep learning

LucianoLalika — Sat, 20 Jun 2020 02:05:11 GMT

Hello everyone,

Currently, I'm working on object detection using deep learning in ArcGIS Pro and the image below is the results I've got. I read a couple of documentations from Esri on how to implement object detection using deep learning in ArcGIS Pro, about the models, parameters, and arguments to be used and I've tried to add datasets and to use different parameters and arguments so as to improve the results but still, I've issues as you see below. Here are my concerns:

How to remove duplicates?
What is the base of selecting values of the arguments i.e. Max Epoch, Batch size, grids, zooms, ratios, and validation during the model training?
What is the base of selecting values of the arguments i.e. padding, threshold, nms_overlap, batch_size, and Max overlap ratio during detection?

topic Improving object detection using deep learning in ArcGIS Pro Questions

Improving object detection using deep learning