Files
ten31-transcripts/Ten31TranscriptsTests/GridCallAnalyzerTests.swift
T
Grant Gilliam a56b47143c Signal: detect the white speaking border (not a coloured one)
Signal's active-speaker cue is a 3px #ffffff rounded border (saturation ≈ 0),
which the saturation-based highlight detector could never see. Per the
Signal-Desktop source review:

- FrameSampler.thinWhitePoints: grid-sample near-white pixels that sit on a
  THIN structure (a non-white pixel within edgeGap on some axis) so a border/
  ring counts but a solid white blob (face, bright video) does not.
- GridCallAnalyzer: combine coloured (saturated) + white (thin) highlight
  pixels; exclude name-text regions so the white footer name can't be mistaken
  for the border; estimate the tile UP from the name footer (nameAtBottom);
  attribute each highlight pixel to exactly one tile by containment (nearest
  centre as tiebreak) so a border can't bleed into an adjacent tile.
- SignalAdapter: white border on, coloured off, name-at-bottom geometry.

Synthetic 4-tile harness now isolates each speaker with no adjacent-tile bleed;
all 15 XCTest cases pass. Real-screenshot geometry calibration still pending.
2026-06-06 09:52:10 -05:00

65 lines
3.2 KiB
Swift

import XCTest
import CoreGraphics
import CoreText
@testable import Ten31Transcripts
/// Validates the visual adapter against synthetic call frames (no real
/// screenshots needed): OCR anchors the tiles and the highlight is attributed to
/// the correct speaker, tracking it as it moves.
final class GridCallAnalyzerTests: XCTestCase {
private func drawText(_ s: String, _ ctx: CGContext, center: CGPoint, size: CGFloat) {
let font = CTFontCreateWithName("Helvetica-Bold" as CFString, size, nil)
let attrs = [kCTFontAttributeName: font,
kCTForegroundColorAttributeName: CGColor(red: 1, green: 1, blue: 1, alpha: 1)] as CFDictionary
let line = CTLineCreateWithAttributedString(CFAttributedStringCreate(nil, s as CFString, attrs)!)
let b = CTLineGetBoundsWithOptions(line, [])
ctx.textPosition = CGPoint(x: center.x - b.width / 2, y: center.y - b.height / 2)
CTLineDraw(line, ctx)
}
private func frame(speakingIndex: Int) -> CGImage {
let W = 800, H = 600
let ctx = CGContext(data: nil, width: W, height: H, bitsPerComponent: 8, bytesPerRow: 0,
space: CGColorSpaceCreateDeviceRGB(),
bitmapInfo: CGImageAlphaInfo.premultipliedLast.rawValue)!
ctx.setFillColor(CGColor(red: 0.1, green: 0.1, blue: 0.12, alpha: 1))
ctx.fill(CGRect(x: 0, y: 0, width: W, height: H))
let rects: [(String, CGRect)] = [
("GRANT", CGRect(x: 40, y: 320, width: 340, height: 230)),
("SARAH", CGRect(x: 420, y: 320, width: 340, height: 230)),
("DMITRI", CGRect(x: 40, y: 50, width: 340, height: 230)),
("ALEX", CGRect(x: 420, y: 50, width: 340, height: 230)),
]
for (i, (name, rect)) in rects.enumerated() {
ctx.setFillColor(CGColor(red: 0.18, green: 0.18, blue: 0.2, alpha: 1)); ctx.fill(rect)
if i == speakingIndex {
// Signal's cue: a WHITE rounded border (not coloured).
ctx.setStrokeColor(CGColor(red: 1, green: 1, blue: 1, alpha: 1)); ctx.setLineWidth(6)
ctx.stroke(rect.insetBy(dx: 3, dy: 3))
}
// Name footer at the BOTTOM of the tile (bottom-left origin: rect.minY).
drawText(name, ctx, center: CGPoint(x: rect.midX, y: rect.minY + 28), size: 46)
}
return ctx.makeImage()!
}
func testReadsNamesAndPicksHighlightedSpeaker() {
let obs = SignalAdapter().analyze(cgImage: frame(speakingIndex: 1), at: 0) // SARAH
XCTAssertGreaterThanOrEqual(obs.count, 2)
let speaking = obs.filter { $0.speaking }
XCTAssertEqual(speaking.count, 1)
// SARAH tile center in top-left pixels (590, 165)
XCTAssertEqual(speaking.first?.bbox.midX ?? 0, 590, accuracy: 160)
XCTAssertEqual(speaking.first?.bbox.midY ?? 0, 165, accuracy: 160)
}
func testHighlightTracksToAnotherTile() {
let obs = SignalAdapter().analyze(cgImage: frame(speakingIndex: 2), at: 1) // DMITRI
let speaking = obs.filter { $0.speaking }
XCTAssertEqual(speaking.count, 1)
XCTAssertEqual(speaking.first?.bbox.midX ?? 0, 210, accuracy: 160)
XCTAssertEqual(speaking.first?.bbox.midY ?? 0, 435, accuracy: 160)
}
}