Snagging Parking Spaces with Mask R-CNN and Python

Step 1: Detecting Parking Spaces in an Image
Here’s what our camera view looks like:
We need be able to scan that image and get back a list of areas that are valid to park in, like this:
Valid parking locations on this city street.The lazy approach would be hardcoding the locations of each parking space into the program by hand instead of trying to detect the parking spaces automatically. Another idea is to build an object detection model that looks for the parking space hash marks drawn on the road, like this:
Notice the tiny yellow marks — those are where the boundaries of each parking space are drawn on the road.But this approach is also painful. If we run the pre-trained model on our camera image, this is what is detected out of the box:

Our image with the default COCO objects identified — cars, people, traffic lights and a tree.

Source: medium.com