Shape Detector

Shape/gesture/stroke detection/recognition algorithm based on the $1 (dollar) recognizer

This is a quick and dirty port of this gesture shape recognition library by MatthieuLoutre to Lua, with some little tweaks. Plays nicely with LÖVE. Should work like the original. If not, please do tell me.

Usage

	local ShapeDetector = require("ShapeDetector")
	local detector = ShapeDetector.new(ShapeDetector.defaultShapes)
	local stroke = {{ x = 127, y = 141 }, { x = 124, y = 140 }, { x = 120, y = 139 }, { x = 118, y = 139 }, { x = 116, y = 139 }, { x = 111, y = 140 }, { x = 109, y = 141 }, { x = 104, y = 144 }, { x = 100, y = 147 }, { x = 96, y = 152 }, { x = 93, y = 157 }, { x = 90, y = 163 }, { x = 87, y = 169 }, { x = 85, y = 175 }, { x = 83, y = 181 }, { x = 82, y = 190 }, { x = 82, y = 195 }, { x = 83, y = 200 }, { x = 84, y = 205 }, { x = 88, y = 213 }, { x = 91, y = 216 }, { x = 96, y = 219 }, { x = 103, y = 222 }, { x = 108, y = 224 }, { x = 111, y = 224 }, { x = 120, y = 224 }, { x = 133, y = 223 }, { x = 142, y = 222 }, { x = 152, y = 218 }, { x = 160, y = 214 }, { x = 167, y = 210 }, { x = 173, y = 204 }, { x = 178, y = 198 }, { x = 179, y = 196 }, { x = 182, y = 188 }, { x = 182, y = 177 }, { x = 178, y = 167 }, { x = 170, y = 150 }, { x = 163, y = 138 }, { x = 152, y = 130 }, { x = 143, y = 129 }, { x = 140, y = 131 }, { x = 129, y = 136 }, { x = 126, y = 139 }}

	-- Just pass a list of points to :spot()!
	local detected = detector:spot(stroke) -- returns circle
	print(detected.pattern)
	print(detected.score)

	-- You can also specify what you're looking for
	detector:spot(stroke, "triangle") -- returns nil as the circle doesn't match the triangle

	-- The detector can also learn new shapes
	local stroke = {{ x = 307, y = 216 }, { x = 333, y = 186 }, { x = 356, y = 215 }, { x = 375, y = 186 }, { x = 399, y = 216 }, { x = 418, y = 186 }}
	detector:learn("zig-zag", stroke)

	-- If the name of multiple shapes start with the same string, the detector will check for all of them
	detector:learn("zig-zag", stroke)
	detector:learn("zig-zag 2", stroke)

	detector:spot(stroke, "zig-zag") -- will check for both "zig-zag"s and for "zig-zag 2" and return the closest

	-- ShapeDetector can also take options
	-- nbSamplePoints (integer) is 64 by default. Increasing it potentially improves accuracy
	-- threshold (0.0-1.0) is 0.9 by default. High numbers are less forgiving to wonky shapes
	-- rotatable is true by default. Allows you to detect a shape regardless of its rotation.
	detector = ShapeDetector.new(ShapeDetector.defaultShapes, { nbSamplePoints = 128, threshold = 0.8, rotatable = true })