Add TPS to project with a "test" showing how to use it

[charbelmarche33]

Changes are made are below:

In the image_registration directory added:

• thin_plate_splining.py: contains a public function that takes src points and some tuning parameters to do TPS
• data/intraop_document_landmarks.json: file that contains the landmarks we are comparing against

In the unit_tests directory added:

• test_tps.py: this simply runs TPS and shows how you would do it... I'm not certain how to ACTUALLY test this.

[RyanDoesMath]

Looks good. I am going to make some changes in the coming week and commit this.

A note on the 4gb of memory being used, memory usage is directly linked to the image's size, and the cv2.remap function is the culprit here. I don't know what this function does under the hood, but apparently it uses a lot of data.

Next semester we should do some experiments to see if decreasing image size impacts accuracy, because if it doesn't we can save a lot of computation time.

[charbelmarche33]

Yeah, that's a great idea. If it's all the same we can shrink the image significantly before calling that method.

That being said, what is the end vision for where this package will end up? Is it going to be run locally on the user's phones or as a microservice in the cloud? Just want to know how cognizant of resource usage we should be, it'll likely help guide some of the work we can do next semester.