Currently we center our app's pipeline around Core Image, which is a high-level image processing library that contains useful out-of-box methods for processing images.
For our use cases, we cannot treat most of our sensor and other input data (depth map, segmentation mask) as image data, and rather as numerical data that should not be tampered with through approximation-based techniques.

Thus, we will have to shift to lower level libraries: Accelerate for CPU-based processing, and Metal/Metal Compute Shaders for GPU-based processing