diff --git a/Sources/Kapture/App.swift b/Sources/Kapture/App.swift
index 939a157..6951f31 100644
--- a/Sources/Kapture/App.swift
+++ b/Sources/Kapture/App.swift
@@ -123,8 +123,8 @@ enum LoginItem {
}
}
-func renderTestImage(_ text: String, fontSize: CGFloat = 56) -> CGImage? {
- let size = NSSize(width: 1900, height: 220)
+func renderTestImage(_ text: String, fontSize: CGFloat = 56, width: CGFloat = 1900, height: CGFloat = 220) -> CGImage? {
+ let size = NSSize(width: width, height: height)
guard let rep = NSBitmapImageRep(
bitmapDataPlanes: nil, pixelsWide: Int(size.width), pixelsHigh: Int(size.height),
bitsPerSample: 8, samplesPerPixel: 4, hasAlpha: true, isPlanar: false,
@@ -138,15 +138,20 @@ func renderTestImage(_ text: String, fontSize: CGFloat = 56) -> CGImage? {
.font: NSFont.systemFont(ofSize: fontSize),
.foregroundColor: NSColor.black,
]
- text.draw(at: NSPoint(x: 60, y: 75), withAttributes: attrs)
+ // Centre the text vertically so tiny canvases still contain it.
+ let textHeight = ("Ag" as NSString).size(withAttributes: attrs).height
+ let y = max(4, (size.height - textHeight) / 2)
+ text.draw(at: NSPoint(x: 20, y: y), withAttributes: attrs)
NSGraphicsContext.restoreGraphicsState()
return rep.cgImage
}
func runOCRSelfTest() {
- let cases: [(expected: String, contains: String)] = [
- ("Kapture OCR test 1234, hello world", "Kapture OCR test 1234"),
- ("你好世界，这是一个测试", "你好世界"),
+ let cases: [(expected: String, contains: String, fontSize: CGFloat, width: CGFloat, height: CGFloat)] = [
+ ("Kapture OCR test 1234, hello world", "Kapture OCR test 1234", 56, 1900, 220),
+ ("你好世界，这是一个测试", "你好世界", 56, 1900, 220),
+ // Tiny text: needs the upscale retry to read anything at all.
+ ("会话导出并发冲突报错 export fail", "导出", 7, 320, 40),
]
var index = 0
@@ -154,9 +159,9 @@ func runOCRSelfTest() {
guard index < cases.count else { exit(0) }
let current = cases[index]
index += 1
- guard let image = renderTestImage(current.expected) else { exit(1) }
+ guard let image = renderTestImage(current.expected, fontSize: current.fontSize, width: current.width, height: current.height) else { exit(1) }
OCRService.shared.recognize(image) { recognized in
- print("case: \(current.expected)")
+ print("case: \(current.expected) @\(Int(current.fontSize))pt")
print("got: \(recognized)")
let ok = recognized.contains(current.contains)
print(ok ? "PASS" : "FAIL")
diff --git a/Sources/Kapture/OCRService.swift b/Sources/Kapture/OCRService.swift
index ca321ff..409c0c1 100644
--- a/Sources/Kapture/OCRService.swift
+++ b/Sources/Kapture/OCRService.swift
@@ -8,34 +8,66 @@ final class OCRService {
func recognize(_ cgImage: CGImage, completion: @escaping (String) -> Void) {
queue.async {
- let request = VNRecognizeTextRequest()
- request.recognitionLevel = .accurate
- request.usesLanguageCorrection = true
- request.recognitionLanguages = ["en-US", "zh-Hans", "zh-Hant", "ja-JP", "ko-KR"]
- // Without this, Vision defaults to the first listed language and
- // renders CJK as Latin garbage (verified: 你好世界 -> "177WR").
- if #available(macOS 13.0, *) {
- request.automaticallyDetectsLanguage = true
- }
-
- let handler = VNImageRequestHandler(cgImage: cgImage, options: [:])
- do {
- try handler.perform([request])
- } catch {
- completion("")
+ let text = Self.perform(cgImage)
+ // Tiny text (small UI labels) often reads as nothing at native
+ // size; retry upscaled and keep whichever pass read more.
+ guard text.count < 3 else {
+ completion(text)
return
}
-
- let observations = request.results ?? []
- let lines = observations
- .sorted { a, b in
- let ay = a.boundingBox.midY
- let by = b.boundingBox.midY
- if abs(ay - by) > 0.02 { return ay > by }
- return a.boundingBox.minX < b.boundingBox.minX
- }
- .compactMap { $0.topCandidates(1).first?.string }
- completion(lines.joined(separator: "\n"))
+ var best = text
+ for factor in [CGFloat(2), CGFloat(3)] {
+ guard let upscaled = Self.upscaled(cgImage, factor: factor) else { continue }
+ let candidate = Self.perform(upscaled)
+ if candidate.count > best.count { best = candidate }
+ if best.count >= 3 { break }
+ }
+ completion(best)
}
}
+
+ private static func perform(_ cgImage: CGImage) -> String {
+ let request = VNRecognizeTextRequest()
+ request.recognitionLevel = .accurate
+ request.usesLanguageCorrection = true
+ request.recognitionLanguages = ["en-US", "zh-Hans", "zh-Hant", "ja-JP", "ko-KR"]
+ // Without this, Vision defaults to the first listed language and
+ // renders CJK as Latin garbage (verified: 你好世界 -> "177WR").
+ if #available(macOS 13.0, *) {
+ request.automaticallyDetectsLanguage = true
+ }
+
+ let handler = VNImageRequestHandler(cgImage: cgImage, options: [:])
+ do {
+ try handler.perform([request])
+ } catch {
+ return ""
+ }
+
+ let observations = request.results ?? []
+ let lines = observations
+ .sorted { a, b in
+ let ay = a.boundingBox.midY
+ let by = b.boundingBox.midY
+ if abs(ay - by) > 0.02 { return ay > by }
+ return a.boundingBox.minX < b.boundingBox.minX
+ }
+ .compactMap { $0.topCandidates(1).first?.string }
+ return lines.joined(separator: "\n")
+ }
+
+ private static func upscaled(_ cgImage: CGImage, factor: CGFloat) -> CGImage? {
+ let width = Int(CGFloat(cgImage.width) * factor)
+ let height = Int(CGFloat(cgImage.height) * factor)
+ guard width > 0, height > 0, width <= 20000, height <= 20000 else { return nil }
+ guard let ctx = CGContext(
+ data: nil, width: width, height: height,
+ bitsPerComponent: 8, bytesPerRow: 0,
+ space: CGColorSpaceCreateDeviceRGB(),
+ bitmapInfo: CGImageAlphaInfo.premultipliedLast.rawValue
+ ) else { return nil }
+ ctx.interpolationQuality = .high
+ ctx.draw(cgImage, in: CGRect(x: 0, y: 0, width: CGFloat(width), height: CGFloat(height)))
+ return ctx.makeImage()
+ }
}
diff --git a/build-app.sh b/build-app.sh
index 564643e..61f6aba 100755
--- a/build-app.sh
+++ b/build-app.sh
@@ -108,9 +108,9 @@ cat > "$APP/Contents/Info.plist" <<'PLIST'
CFBundleIconFile
Icon
CFBundleShortVersionString
- 1.10.1
+ 1.10.2
CFBundleVersion
- 101
+ 102
LSUIElement
NSHighResolutionCapable