ColourShape benchmark and moar docs (#11)

* Added benchmark

* Rebase

* [Server] ColouredShapePipeline Benchmarks and Documentation

* [Server] ColouredShapePipeline Benchmarks and Documentation

* Added benchmark

* Rebase

* [Server] ColouredShapePipeline Benchmarks and Documentation

* [Server] ColouredShapePipeline Benchmarks and Documentation

* [Server] ColouredShapePipeline Benchmarks and Documentation

* [CSP] Rebase off master

* [CSP] Remove unused variables

* [CSP] Make circles Mat private and final
This commit is contained in:
Xzibit
2020-07-12 14:36:45 -04:00
committed by GitHub
parent 54445dad35
commit 7a7f2ff91c
4 changed files with 274 additions and 6 deletions

View File

@@ -31,12 +31,17 @@ import org.photonvision.vision.pipe.CVPipe;
public class FindCirclesPipe
extends CVPipe<Pair<Mat, List<Contour>>, List<CVShape>, FindCirclesPipe.FindCirclePipeParams> {
// Output vector of found circles. Each vector is encoded as 3 or 4 element floating-point vector
// (x,y,radius) or (x,y,radius,votes) .
private final Mat circles = new Mat();
/**
* Runs the process for the pipe.
* Runs the process for the pipe. The reason we need a separate pipe for circles is because if we
* were to use the FindShapes pipe, we would have to assume that any shape more than 10-20+ sides
* is a circle. Only issue with such approximation is that the user would no longer be able to
* track shapes with 10-20+ sides. And hence, in order to overcome this edge case, we can use
* HoughCircles which is more flexible and accurate for finding circles.
*
* @param in Input for pipe processing.
* @param in Input for pipe processing. 8-bit, single-channel, grayscale input image.
* @return Result of processing.
*/
@Override
@@ -47,27 +52,40 @@ public class FindCirclesPipe
Imgproc.HoughCircles(
in.getLeft(),
circles,
// Detection method, see #HoughModes. The available methods are #HOUGH_GRADIENT and
// #HOUGH_GRADIENT_ALT.
Imgproc.HOUGH_GRADIENT,
/*Inverse ratio of the accumulator resolution to the image resolution.
For example, if dp=1 , the accumulator has the same resolution as the input image.
If dp=2 , the accumulator has half as big width and height. For #HOUGH_GRADIENT_ALT the recommended value is dp=1.5,
unless some small very circles need to be detected.
*/
1.0,
params.minDist,
params.maxCannyThresh,
params.accuracy,
params.minRadius,
params.maxRadius);
// Great, we now found the center point of the circle and it's radius, but we have no idea what
// contour it corresponds to
for (int x = 0; x < circles.cols(); x++) {
// Grab the current circle we are looking at
double[] c = circles.get(0, x);
// Find the center points of that circle
double x_center = c[0];
double y_center = c[1];
for (Contour contour : in.getRight()) {
// Grab the moments of the current contour
Moments mu = contour.getMoments();
// Determine if the contour is within the threshold of the detected circle
if (Math.abs(x_center - (mu.m10 / mu.m00)) <= params.allowableThreshold
&& Math.abs(y_center - (mu.m01 / mu.m00)) <= params.allowableThreshold) {
// If it is, then add it to the output array
output.add(new CVShape(contour, ContourShape.Circle));
}
}
}
return output;
}
@@ -79,6 +97,17 @@ public class FindCirclesPipe
private final int maxCannyThresh;
private final int accuracy;
/*
* @params minDist - Minimum distance between the centers of the detected circles.
* If the parameter is too small, multiple neighbor circles may be falsely detected in addition to a true one. If it is too large, some circles may be missed.
*
* @param maxCannyThresh -First method-specific parameter. In case of #HOUGH_GRADIENT and #HOUGH_GRADIENT_ALT, it is the higher threshold of the two passed to the Canny edge detector (the lower one is twice smaller).
* Note that #HOUGH_GRADIENT_ALT uses #Scharr algorithm to compute image derivatives, so the threshold value shough normally be higher, such as 300 or normally exposed and contrasty images.
*
*
* @param allowableThreshold - When finding the corresponding contour, this is used to see how close a center should be to a contour for it to be considered THAT contour.
* Should be increased with lower resolutions and decreased with higher resolution
* */
public FindCirclePipeParams(
int allowableThreshold,
int minRadius,

View File

@@ -49,6 +49,11 @@ public class FindPolygonPipe
private CVShape getShape(Contour in) {
int corners = getCorners(in);
corners = getCorners(in);
/*The contourShape enum has predefined shapes for Circles, Triangles, and Quads
meaning any shape not fitting in those predefined shapes must be a custom shape.
*/
if (ContourShape.fromSides(corners) == null) {
return new CVShape(in, ContourShape.Custom);
}
@@ -65,18 +70,21 @@ public class FindPolygonPipe
}
private int getCorners(Contour contour) {
// Release previous approx
approx.release();
Imgproc.approxPolyDP(
contour.getMat2f(),
approx,
// Converts an accuracy percentage between 1-100 to an epsilon
params.accuracyPercentage / 600.0 * Imgproc.arcLength(contour.getMat2f(), true),
true);
// The height of the resultant approximation is the number of vertices
return (int) approx.size().height;
}
public static class FindPolygonPipeParams {
private final double accuracyPercentage;
// Should be a value between 0-100
public FindPolygonPipeParams(double accuracyPercentage) {
this.accuracyPercentage = accuracyPercentage;
}

View File

@@ -61,7 +61,7 @@ public class ColoredShapePipeline
private CVPipeResult<List<TrackedTarget>> targetList;
private final Point[] rectPoints = new Point[4];
ColoredShapePipeline() {
public ColoredShapePipeline() {
settings = new ColoredShapePipelineSettings();
}

View File

@@ -0,0 +1,231 @@
/*
* Copyright (C) 2020 Photon Vision.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <https://www.gnu.org/licenses/>.
*/
package org.photonvision.common;
/*
* Copyright (C) 2020 Photon Vision.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <https://www.gnu.org/licenses/>.
*/
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import org.junit.jupiter.api.BeforeAll;
import org.junit.jupiter.api.Test;
import org.photonvision.common.util.TestUtils;
import org.photonvision.common.util.math.MathUtils;
import org.photonvision.common.util.numbers.NumberListUtils;
import org.photonvision.vision.frame.FrameProvider;
import org.photonvision.vision.frame.provider.FileFrameProvider;
import org.photonvision.vision.opencv.CVMat;
import org.photonvision.vision.opencv.ContourGroupingMode;
import org.photonvision.vision.opencv.ContourIntersectionDirection;
import org.photonvision.vision.opencv.ContourShape;
import org.photonvision.vision.pipeline.CVPipeline;
import org.photonvision.vision.pipeline.ColoredShapePipeline;
import org.photonvision.vision.pipeline.result.CVPipelineResult;
/** Various tests that check performance on long-running tasks (i.e. a pipeline) */
public class ShapeBenchmarkTest {
@BeforeAll
public static void init() {
TestUtils.loadLibraries();
}
@Test
public void Shape240pBenchmark() {
var pipeline = new ColoredShapePipeline();
pipeline.getSettings().hsvHue.set(60, 100);
pipeline.getSettings().hsvSaturation.set(100, 255);
pipeline.getSettings().hsvValue.set(190, 255);
pipeline.getSettings().outputShowThresholded = true;
pipeline.getSettings().outputShowMultipleTargets = true;
pipeline.getSettings().contourGroupingMode = ContourGroupingMode.Single;
pipeline.getSettings().contourIntersection = ContourIntersectionDirection.Up;
pipeline.getSettings().desiredShape = ContourShape.Custom;
pipeline.getSettings().allowableThreshold = 10;
pipeline.getSettings().accuracyPercentage = 30.0;
var frameProvider =
new FileFrameProvider(
TestUtils.getWPIImagePath(TestUtils.WPI2019Image.kCargoSideStraightDark72in),
TestUtils.WPI2019Image.FOV);
frameProvider.setImageReloading(true);
benchmarkPipeline(frameProvider, pipeline, 5);
}
@Test
public void Shape480pBenchmark() {
var pipeline = new ColoredShapePipeline();
pipeline.getSettings().hsvHue.set(60, 100);
pipeline.getSettings().hsvSaturation.set(100, 255);
pipeline.getSettings().hsvValue.set(190, 255);
pipeline.getSettings().outputShowThresholded = true;
pipeline.getSettings().outputShowMultipleTargets = true;
pipeline.getSettings().contourGroupingMode = ContourGroupingMode.Single;
pipeline.getSettings().contourIntersection = ContourIntersectionDirection.Up;
pipeline.getSettings().desiredShape = ContourShape.Custom;
pipeline.getSettings().allowableThreshold = 10;
pipeline.getSettings().accuracyPercentage = 30.0;
var frameProvider =
new FileFrameProvider(
TestUtils.getWPIImagePath(TestUtils.WPI2020Image.kBlueGoal_084in_Center),
TestUtils.WPI2020Image.FOV);
frameProvider.setImageReloading(true);
benchmarkPipeline(frameProvider, pipeline, 5);
}
@Test
public void Shape720pBenchmark() {
var pipeline = new ColoredShapePipeline();
pipeline.getSettings().hsvHue.set(60, 100);
pipeline.getSettings().hsvSaturation.set(100, 255);
pipeline.getSettings().hsvValue.set(190, 255);
pipeline.getSettings().outputShowThresholded = true;
pipeline.getSettings().outputShowMultipleTargets = true;
pipeline.getSettings().contourGroupingMode = ContourGroupingMode.Single;
pipeline.getSettings().contourIntersection = ContourIntersectionDirection.Up;
pipeline.getSettings().desiredShape = ContourShape.Custom;
pipeline.getSettings().allowableThreshold = 10;
pipeline.getSettings().accuracyPercentage = 30.0;
var frameProvider =
new FileFrameProvider(
TestUtils.getWPIImagePath(TestUtils.WPI2020Image.kBlueGoal_084in_Center_720p),
TestUtils.WPI2020Image.FOV);
frameProvider.setImageReloading(true);
benchmarkPipeline(frameProvider, pipeline, 5);
}
@Test
public void Shape1920x1440Benchmark() {
var pipeline = new ColoredShapePipeline();
pipeline.getSettings().hsvHue.set(60, 100);
pipeline.getSettings().hsvSaturation.set(100, 255);
pipeline.getSettings().hsvValue.set(190, 255);
pipeline.getSettings().outputShowThresholded = true;
pipeline.getSettings().outputShowMultipleTargets = true;
pipeline.getSettings().contourGroupingMode = ContourGroupingMode.Single;
pipeline.getSettings().contourIntersection = ContourIntersectionDirection.Up;
pipeline.getSettings().desiredShape = ContourShape.Custom;
pipeline.getSettings().allowableThreshold = 10;
pipeline.getSettings().accuracyPercentage = 30.0;
var frameProvider =
new FileFrameProvider(
TestUtils.getWPIImagePath(TestUtils.WPI2019Image.kCargoStraightDark72in_HighRes),
TestUtils.WPI2019Image.FOV);
frameProvider.setImageReloading(true);
benchmarkPipeline(frameProvider, pipeline, 5);
}
private static <P extends CVPipeline> void benchmarkPipeline(
FrameProvider frameProvider, P pipeline, int secondsToRun) {
CVMat.enablePrint(false);
// warmup for 5 loops.
System.out.println("Warming up for 5 loops...");
for (int i = 0; i < 5; i++) {
pipeline.run(frameProvider.get());
}
final List<Double> processingTimes = new ArrayList<>();
final List<Double> latencyTimes = new ArrayList<>();
var frameProps = frameProvider.get().frameStaticProperties;
// begin benchmark
System.out.println(
"Beginning "
+ secondsToRun
+ " second benchmark at resolution "
+ frameProps.imageWidth
+ "x"
+ frameProps.imageHeight);
var benchmarkStartMillis = System.currentTimeMillis();
do {
CVPipelineResult pipelineResult = pipeline.run(frameProvider.get());
pipelineResult.release();
processingTimes.add(pipelineResult.processingMillis);
latencyTimes.add(pipelineResult.getLatencyMillis());
} while (System.currentTimeMillis() - benchmarkStartMillis < secondsToRun * 1000);
System.out.println("Benchmark complete.");
var processingMin = Collections.min(processingTimes);
var processingMean = NumberListUtils.mean(processingTimes);
var processingMax = Collections.max(processingTimes);
var latencyMin = Collections.min(latencyTimes);
var latencyMean = NumberListUtils.mean(latencyTimes);
var latencyMax = Collections.max(latencyTimes);
String processingResult =
"Processing times - "
+ "Min: "
+ MathUtils.roundTo(processingMin, 3)
+ "ms ("
+ MathUtils.roundTo(1000 / processingMin, 3)
+ " FPS), "
+ "Mean: "
+ MathUtils.roundTo(processingMean, 3)
+ "ms ("
+ MathUtils.roundTo(1000 / processingMean, 3)
+ " FPS), "
+ "Max: "
+ MathUtils.roundTo(processingMax, 3)
+ "ms ("
+ MathUtils.roundTo(1000 / processingMax, 3)
+ " FPS)";
System.out.println(processingResult);
String latencyResult =
"Latency times - "
+ "Min: "
+ MathUtils.roundTo(latencyMin, 3)
+ "ms ("
+ MathUtils.roundTo(1000 / latencyMin, 3)
+ " FPS), "
+ "Mean: "
+ MathUtils.roundTo(latencyMean, 3)
+ "ms ("
+ MathUtils.roundTo(1000 / latencyMean, 3)
+ " FPS), "
+ "Max: "
+ MathUtils.roundTo(latencyMax, 3)
+ "ms ("
+ MathUtils.roundTo(1000 / latencyMax, 3)
+ " FPS)";
System.out.println(latencyResult);
}
}