R-CNN in R for analyzing aerial photos

Question

I have 8,000 drone photos on which we want to detect almost circular holes in the ground (no UFOs..). I have been reading and practicing quite a lot with using CNNs in R, using both Keras and Tensorflow. However, I think what I need is a R-CNN instead of a conventional CNN, to be able to scan through the photos looking for the object.

But is there any way to work with R-CNN or even Mask R-CNN in R? I have no experience with Python.

Answer 1

As you are familiar with using Keras and Tensorflow in R, you should be able to implement the architecture as per your needs. You will be able to find the rough architecture of most models with a simple Google search and then tweak it to fit your use case.

For example - Faster RCNN architecture.