Mike/Gaze
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

lol

Browse Source
This commit is contained in:
Michael Freno
2026-01-27 12:50:23 -05:00
parent b1996c8324
commit a5602d9829
9 changed files with 198 additions and 139 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
GazeTests/Services/PupilDetectorTests.swift
View File

@@ -2,41 +2,42 @@
// PupilDetectorTests.swift
// GazeTests
//
// Created by Claude on 1/16/26.
// Created by Mike Freno on 1/16/26.
//
import XCTest
import CoreVideo
import Vision
import XCTest
@testable import Gaze
final class PupilDetectorTests: XCTestCase {
override func setUp() async throws {
// Reset the detector state
PupilDetector.cleanup()
}
func testCreateCGImageFromData() throws {
// Test basic image creation
let width = 50
let height = 50
var pixels = [UInt8](repeating: 128, count: width * height)
// Add some dark pixels for a "pupil"
for y in 20..<30 {
for x in 20..<30 {
pixels[y * width + x] = 10 // Very dark
}
}
// Save test image to verify
let pixelData = Data(pixels)
guard let provider = CGDataProvider(data: pixelData as CFData) else {
XCTFail("Failed to create CGDataProvider")
return
}
let cgImage = CGImage(
width: width,
height: height,
@@ -50,18 +51,18 @@ final class PupilDetectorTests: XCTestCase {
shouldInterpolate: false,
intent: .defaultIntent
)
XCTAssertNotNil(cgImage, "Should create CGImage from pixel data")
}
func testImageProcessingWithDarkPixels() throws {
// Test that imageProcessingOptimized produces dark pixels
let width = 60
let height = 40
// Create input with a dark circle (simulating pupil)
var input = [UInt8](repeating: 200, count: width * height) // Light background (like eye white)
// Add a dark ellipse in center (pupil)
let centerX = width / 2
let centerY = height / 2
@@ -74,10 +75,10 @@ final class PupilDetectorTests: XCTestCase {
}
}
}
var output = [UInt8](repeating: 255, count: width * height)
let threshold = 50 // Same as default
// Call the actual processing function
input.withUnsafeMutableBufferPointer { inputPtr in
output.withUnsafeMutableBufferPointer { outputPtr in
@@ -85,7 +86,7 @@ final class PupilDetectorTests: XCTestCase {
// But we can verify by saving input for inspection
}
}
// Save the input for manual inspection
let inputData = Data(input)
let url = URL(fileURLWithPath: "/Users/mike/gaze/images/test_input_synthetic.png")
@@ -103,28 +104,30 @@ final class PupilDetectorTests: XCTestCase {
shouldInterpolate: false,
intent: .defaultIntent
) {
if let dest = CGImageDestinationCreateWithURL(url as CFURL, "public.png" as CFString, 1, nil) {
if let dest = CGImageDestinationCreateWithURL(
url as CFURL, "public.png" as CFString, 1, nil)
{
CGImageDestinationAddImage(dest, cgImage, nil)
CGImageDestinationFinalize(dest)
print("💾 Saved synthetic test input to: \(url.path)")
}
}
}
// Count dark pixels in input
let darkCount = input.filter { $0 < 50 }.count
print("📊 Input has \(darkCount) dark pixels (< 50)")
XCTAssertGreaterThan(darkCount, 0, "Input should have dark pixels for pupil")
}
func testFindPupilFromContoursWithSyntheticData() throws {
// Create synthetic binary image with a dark region
let width = 60
let height = 40
// All white except a dark blob
var binaryData = [UInt8](repeating: 255, count: width * height)
// Add dark region (0 = dark/pupil)
let centerX = 30
let centerY = 20
@@ -139,9 +142,9 @@ final class PupilDetectorTests: XCTestCase {
}
}
}
print("📊 Created synthetic image with \(darkPixelCount) dark pixels")
// Save for inspection
let binaryDataObj = Data(binaryData)
let url = URL(fileURLWithPath: "/Users/mike/gaze/images/test_binary_synthetic.png")
@@ -159,14 +162,16 @@ final class PupilDetectorTests: XCTestCase {
shouldInterpolate: false,
intent: .defaultIntent
) {
if let dest = CGImageDestinationCreateWithURL(url as CFURL, "public.png" as CFString, 1, nil) {
if let dest = CGImageDestinationCreateWithURL(
url as CFURL, "public.png" as CFString, 1, nil)
{
CGImageDestinationAddImage(dest, cgImage, nil)
CGImageDestinationFinalize(dest)
print("💾 Saved synthetic binary image to: \(url.path)")
}
}
}
XCTAssertGreaterThan(darkPixelCount, 10, "Should have enough dark pixels")
}
}

229
GazeTests/VideoGazeTests.swift
View File

@@ -2,77 +2,103 @@
// VideoGazeTests.swift
// GazeTests
//
// Created by Claude on 1/16/26.
// Created by Mike Freno on 1/16/26.
//
import XCTest
import AVFoundation
import Vision
import XCTest
@testable import Gaze
final class VideoGazeTests: XCTestCase {
var logLines: [String] = []
private func log(_ message: String) {
logLines.append(message)
}
private func attachLogs() {
let attachment = XCTAttachment(string: logLines.joined(separator: "\n"))
attachment.name = "Test Logs"
attachment.lifetime = .keepAlways
add(attachment)
}
/// Process the outer video (looking away from screen) - should detect "looking away"
func testOuterVideoGazeDetection() async throws {
logLines = []
let projectPath = "/Users/mike/Code/Gaze/GazeTests/video-test-outer.mp4"
guard FileManager.default.fileExists(atPath: projectPath) else {
XCTFail("Video file not found at: \(projectPath)")
return
}
let stats = try await processVideo(at: URL(fileURLWithPath: projectPath), expectLookingAway: true)
let stats = try await processVideo(
at: URL(fileURLWithPath: projectPath), expectLookingAway: true)
// For outer video, most frames should detect gaze outside center
let nonCenterRatio = Double(stats.nonCenterFrames) / Double(max(1, stats.pupilDetectedFrames))
log("🎯 OUTER video: \(String(format: "%.1f%%", nonCenterRatio * 100)) frames detected as non-center (expected: >50%)")
log(" H-range: \(String(format: "%.3f", stats.minH)) to \(String(format: "%.3f", stats.maxH))")
log(" V-range: \(String(format: "%.3f", stats.minV)) to \(String(format: "%.3f", stats.maxV))")
log(" Face width: \(String(format: "%.3f", stats.avgFaceWidth)) (range: \(String(format: "%.3f", stats.minFaceWidth))-\(String(format: "%.3f", stats.maxFaceWidth)))")
let nonCenterRatio =
Double(stats.nonCenterFrames) / Double(max(1, stats.pupilDetectedFrames))
log(
"🎯 OUTER video: \(String(format: "%.1f%%", nonCenterRatio * 100)) frames detected as non-center (expected: >50%)"
)
log(
" H-range: \(String(format: "%.3f", stats.minH)) to \(String(format: "%.3f", stats.maxH))"
)
log(
" V-range: \(String(format: "%.3f", stats.minV)) to \(String(format: "%.3f", stats.maxV))"
)
log(
" Face width: \(String(format: "%.3f", stats.avgFaceWidth)) (range: \(String(format: "%.3f", stats.minFaceWidth))-\(String(format: "%.3f", stats.maxFaceWidth)))"
)
attachLogs()
// At least 50% should be detected as non-center when looking away
XCTAssertGreaterThan(nonCenterRatio, 0.5, "Looking away video should have >50% non-center detections. Log:\n\(logLines.joined(separator: "\n"))")
XCTAssertGreaterThan(
nonCenterRatio, 0.5,
"Looking away video should have >50% non-center detections. Log:\n\(logLines.joined(separator: "\n"))"
)
}
/// Process the inner video (looking at screen) - should detect "looking at screen"
func testInnerVideoGazeDetection() async throws {
logLines = []
let projectPath = "/Users/mike/Code/Gaze/GazeTests/video-test-inner.mp4"
guard FileManager.default.fileExists(atPath: projectPath) else {
XCTFail("Video file not found at: \(projectPath)")
return
}
let stats = try await processVideo(at: URL(fileURLWithPath: projectPath), expectLookingAway: false)
let stats = try await processVideo(
at: URL(fileURLWithPath: projectPath), expectLookingAway: false)
// For inner video, most frames should detect gaze at center
let centerRatio = Double(stats.centerFrames) / Double(max(1, stats.pupilDetectedFrames))
log("🎯 INNER video: \(String(format: "%.1f%%", centerRatio * 100)) frames detected as center (expected: >50%)")
log(" H-range: \(String(format: "%.3f", stats.minH)) to \(String(format: "%.3f", stats.maxH))")
log(" V-range: \(String(format: "%.3f", stats.minV)) to \(String(format: "%.3f", stats.maxV))")
log(" Face width: \(String(format: "%.3f", stats.avgFaceWidth)) (range: \(String(format: "%.3f", stats.minFaceWidth))-\(String(format: "%.3f", stats.maxFaceWidth)))")
log(
"🎯 INNER video: \(String(format: "%.1f%%", centerRatio * 100)) frames detected as center (expected: >50%)"
)
log(
" H-range: \(String(format: "%.3f", stats.minH)) to \(String(format: "%.3f", stats.maxH))"
)
log(
" V-range: \(String(format: "%.3f", stats.minV)) to \(String(format: "%.3f", stats.maxV))"
)
log(
" Face width: \(String(format: "%.3f", stats.avgFaceWidth)) (range: \(String(format: "%.3f", stats.minFaceWidth))-\(String(format: "%.3f", stats.maxFaceWidth)))"
)
attachLogs()
// At least 50% should be detected as center when looking at screen
XCTAssertGreaterThan(centerRatio, 0.5, "Looking at screen video should have >50% center detections. Log:\n\(logLines.joined(separator: "\n"))")
XCTAssertGreaterThan(
centerRatio, 0.5,
"Looking at screen video should have >50% center detections. Log:\n\(logLines.joined(separator: "\n"))"
)
}
struct VideoStats {
var totalFrames = 0
var faceDetectedFrames = 0
@@ -87,34 +113,38 @@ final class VideoGazeTests: XCTestCase {
var maxFaceWidth = -Double.greatestFiniteMagnitude
var totalFaceWidth = 0.0
var faceWidthCount = 0
var avgFaceWidth: Double {
faceWidthCount > 0 ? totalFaceWidth / Double(faceWidthCount) : 0
}
}
private func processVideo(at url: URL, expectLookingAway: Bool) async throws -> VideoStats {
var stats = VideoStats()
log("\n" + String(repeating: "=", count: 60))
log("Processing video: \(url.lastPathComponent)")
log("Expected behavior: \(expectLookingAway ? "LOOKING AWAY (non-center)" : "LOOKING AT SCREEN (center)")")
log(
"Expected behavior: \(expectLookingAway ? "LOOKING AWAY (non-center)" : "LOOKING AT SCREEN (center)")"
)
log(String(repeating: "=", count: 60))
let asset = AVURLAsset(url: url)
let duration = try await asset.load(.duration)
let durationSeconds = CMTimeGetSeconds(duration)
log("Duration: \(String(format: "%.2f", durationSeconds)) seconds")
guard let track = try await asset.loadTracks(withMediaType: .video).first else {
XCTFail("No video track found")
return stats
}
let size = try await track.load(.naturalSize)
let frameRate = try await track.load(.nominalFrameRate)
log("Size: \(Int(size.width))x\(Int(size.height)), FPS: \(String(format: "%.1f", frameRate))")
log(
"Size: \(Int(size.width))x\(Int(size.height)), FPS: \(String(format: "%.1f", frameRate))"
)
let reader = try AVAssetReader(asset: asset)
let outputSettings: [String: Any] = [
kCVPixelBufferPixelFormatTypeKey as String: kCVPixelFormatType_32BGRA
@@ -122,82 +152,86 @@ final class VideoGazeTests: XCTestCase {
let trackOutput = AVAssetReaderTrackOutput(track: track, outputSettings: outputSettings)
reader.add(trackOutput)
reader.startReading()
var frameIndex = 0
let sampleInterval = max(1, Int(frameRate / 2)) // Sample ~2 frames per second
let sampleInterval = max(1, Int(frameRate / 2)) // Sample ~2 frames per second
log("\nFrame | Time | Face | H-Ratio L/R | V-Ratio L/R | Direction")
log(String(repeating: "-", count: 75))
// Reset calibration for fresh test
PupilDetector.calibration.reset()
// Disable frame skipping for video testing
let originalFrameSkip = PupilDetector.frameSkipCount
PupilDetector.frameSkipCount = 1
defer { PupilDetector.frameSkipCount = originalFrameSkip }
while let sampleBuffer = trackOutput.copyNextSampleBuffer() {
defer {
frameIndex += 1
defer {
frameIndex += 1
PupilDetector.advanceFrame()
}
// Only process every Nth frame
if frameIndex % sampleInterval != 0 {
continue
}
stats.totalFrames += 1
guard let pixelBuffer = CMSampleBufferGetImageBuffer(sampleBuffer) else {
continue
}
let timestamp = CMSampleBufferGetPresentationTimeStamp(sampleBuffer)
let timeSeconds = CMTimeGetSeconds(timestamp)
// Run face detection
let request = VNDetectFaceLandmarksRequest()
request.revision = VNDetectFaceLandmarksRequestRevision3
let handler = VNImageRequestHandler(
cvPixelBuffer: pixelBuffer,
orientation: .leftMirrored,
options: [:]
)
try handler.perform([request])
guard let observations = request.results, !observations.isEmpty,
let face = observations.first,
let landmarks = face.landmarks,
let leftEye = landmarks.leftEye,
let rightEye = landmarks.rightEye else {
log(String(format: "%5d | %5.1fs | NO | - | - | -", frameIndex, timeSeconds))
let face = observations.first,
let landmarks = face.landmarks,
let leftEye = landmarks.leftEye,
let rightEye = landmarks.rightEye
else {
log(
String(
format: "%5d | %5.1fs | NO | - | - | -",
frameIndex, timeSeconds))
continue
}
stats.faceDetectedFrames += 1
// Track face width (bounding box width as ratio of image width)
let faceWidth = face.boundingBox.width
stats.minFaceWidth = min(stats.minFaceWidth, faceWidth)
stats.maxFaceWidth = max(stats.maxFaceWidth, faceWidth)
stats.totalFaceWidth += faceWidth
stats.faceWidthCount += 1
let imageSize = CGSize(
width: CVPixelBufferGetWidth(pixelBuffer),
height: CVPixelBufferGetHeight(pixelBuffer)
)
// Detect pupils
var leftHRatio: Double?
var rightHRatio: Double?
var leftVRatio: Double?
var rightVRatio: Double?
if let leftResult = PupilDetector.detectPupil(
in: pixelBuffer,
eyeLandmarks: leftEye,
@@ -205,10 +239,12 @@ final class VideoGazeTests: XCTestCase {
imageSize: imageSize,
side: 0
) {
leftHRatio = calculateHorizontalRatio(pupilPosition: leftResult.pupilPosition, eyeRegion: leftResult.eyeRegion)
leftVRatio = calculateVerticalRatio(pupilPosition: leftResult.pupilPosition, eyeRegion: leftResult.eyeRegion)
leftHRatio = calculateHorizontalRatio(
pupilPosition: leftResult.pupilPosition, eyeRegion: leftResult.eyeRegion)
leftVRatio = calculateVerticalRatio(
pupilPosition: leftResult.pupilPosition, eyeRegion: leftResult.eyeRegion)
}
if let rightResult = PupilDetector.detectPupil(
in: pixelBuffer,
eyeLandmarks: rightEye,
@@ -216,62 +252,79 @@ final class VideoGazeTests: XCTestCase {
imageSize: imageSize,
side: 1
) {
rightHRatio = calculateHorizontalRatio(pupilPosition: rightResult.pupilPosition, eyeRegion: rightResult.eyeRegion)
rightVRatio = calculateVerticalRatio(pupilPosition: rightResult.pupilPosition, eyeRegion: rightResult.eyeRegion)
rightHRatio = calculateHorizontalRatio(
pupilPosition: rightResult.pupilPosition, eyeRegion: rightResult.eyeRegion)
rightVRatio = calculateVerticalRatio(
pupilPosition: rightResult.pupilPosition, eyeRegion: rightResult.eyeRegion)
}
if let lh = leftHRatio, let rh = rightHRatio,
let lv = leftVRatio, let rv = rightVRatio {
let lv = leftVRatio, let rv = rightVRatio
{
stats.pupilDetectedFrames += 1
let avgH = (lh + rh) / 2.0
let avgV = (lv + rv) / 2.0
// Track min/max ranges
stats.minH = min(stats.minH, avgH)
stats.maxH = max(stats.maxH, avgH)
stats.minV = min(stats.minV, avgV)
stats.maxV = max(stats.maxV, avgV)
let direction = GazeDirection.from(horizontal: avgH, vertical: avgV)
if direction == .center {
stats.centerFrames += 1
} else {
stats.nonCenterFrames += 1
}
log(String(format: "%5d | %5.1fs | YES | %.2f / %.2f | %.2f / %.2f | %@ %@",
frameIndex, timeSeconds, lh, rh, lv, rv, direction.rawValue, String(describing: direction)))
log(
String(
format: "%5d | %5.1fs | YES | %.2f / %.2f | %.2f / %.2f | %@ %@",
frameIndex, timeSeconds, lh, rh, lv, rv, direction.rawValue,
String(describing: direction)))
} else {
log(String(format: "%5d | %5.1fs | YES | PUPIL FAIL | PUPIL FAIL | -", frameIndex, timeSeconds))
log(
String(
format: "%5d | %5.1fs | YES | PUPIL FAIL | PUPIL FAIL | -",
frameIndex, timeSeconds))
}
}
log(String(repeating: "=", count: 75))
log("Summary: \(stats.totalFrames) frames sampled, \(stats.faceDetectedFrames) with face, \(stats.pupilDetectedFrames) with pupils")
log(
"Summary: \(stats.totalFrames) frames sampled, \(stats.faceDetectedFrames) with face, \(stats.pupilDetectedFrames) with pupils"
)
log("Center frames: \(stats.centerFrames), Non-center: \(stats.nonCenterFrames)")
log("Face width: avg=\(String(format: "%.3f", stats.avgFaceWidth)), range=\(String(format: "%.3f", stats.minFaceWidth)) to \(String(format: "%.3f", stats.maxFaceWidth))")
log(
"Face width: avg=\(String(format: "%.3f", stats.avgFaceWidth)), range=\(String(format: "%.3f", stats.minFaceWidth)) to \(String(format: "%.3f", stats.maxFaceWidth))"
)
log("Processing complete\n")
return stats
}
private func calculateHorizontalRatio(pupilPosition: PupilPosition, eyeRegion: EyeRegion) -> Double {
private func calculateHorizontalRatio(pupilPosition: PupilPosition, eyeRegion: EyeRegion)
-> Double
{
// pupilPosition.y controls horizontal gaze due to image orientation
let pupilY = Double(pupilPosition.y)
let eyeHeight = Double(eyeRegion.frame.height)
guard eyeHeight > 0 else { return 0.5 }
let ratio = pupilY / eyeHeight
return max(0.0, min(1.0, ratio))
}
private func calculateVerticalRatio(pupilPosition: PupilPosition, eyeRegion: EyeRegion) -> Double {
private func calculateVerticalRatio(pupilPosition: PupilPosition, eyeRegion: EyeRegion)
-> Double
{
// pupilPosition.x controls vertical gaze due to image orientation
let pupilX = Double(pupilPosition.x)
let eyeWidth = Double(eyeRegion.frame.width)
guard eyeWidth > 0 else { return 0.5 }
let ratio = pupilX / eyeWidth
return max(0.0, min(1.0, ratio))
}