Depth Map Calculation: In DepthMapProcessor, use something like trimmed mean to eliminate outliers, instead of the normal mean

[himanshunaidu]

Will have to create a separate method for this, and replace the original getDepth and calcLocation methods:

iOSPointMapper/IOSAccessAssessment/Location/ObjectLocation.swift

Lines 102 to 107 in b93a57e

	// FIXME: Use something like trimmed mean to eliminate outliers, instead of the normal mean
	func getDepth(sharedImageData: SharedImageData, index: Int) {
	let objectSegmentation = sharedImageData.classImages[index]
	let mask = createMask(from: objectSegmentation)
	guard let depthMap = sharedImageData.depthData else { return }
	// var distanceSum: Float = 0

[himanshunaidu]

Implementation of the trimmed mean should ideally be done in the GPU using Metal, due to the high number of depth pixels that will have to be processed.

Algorithm Outline:

1. Create a Metal shader to filter and extract valid depths (invalid depths would have a value of 0)
2. Create a Metal shader to move the valid depth values to a (global) float array (can be combined with Step 1)
3. Create a Metal shader to perform mean and standard deviation calculation ignoring the 0 values
4. Create a Metal shader to remove values that are 2 standard deviations away from the mean
5. Create a Metal shader to calculate the final mean ignoring the 0 values

Step 3-5 can probably be efficiently calculated using frameworks like Accelerate as well, as long as there is no plan to do this in real-time (which may eventually be a plan)