this repo has no description
can recursively loop through a directory and output json
+49
-67
+4
-2
OCR.xcodeproj/project.pbxproj
+4
-2
OCR.xcodeproj/project.pbxproj
···
178
178
GCC_WARN_UNINITIALIZED_AUTOS = YES_AGGRESSIVE;
179
179
GCC_WARN_UNUSED_FUNCTION = YES;
180
180
GCC_WARN_UNUSED_VARIABLE = YES;
181
-
MACOSX_DEPLOYMENT_TARGET = 12.0;
181
+
MACOSX_DEPLOYMENT_TARGET = 14.0;
182
182
MTL_ENABLE_DEBUG_INFO = INCLUDE_SOURCE;
183
183
MTL_FAST_MATH = YES;
184
184
ONLY_ACTIVE_ARCH = YES;
···
233
233
GCC_WARN_UNINITIALIZED_AUTOS = YES_AGGRESSIVE;
234
234
GCC_WARN_UNUSED_FUNCTION = YES;
235
235
GCC_WARN_UNUSED_VARIABLE = YES;
236
-
MACOSX_DEPLOYMENT_TARGET = 12.0;
236
+
MACOSX_DEPLOYMENT_TARGET = 14.0;
237
237
MTL_ENABLE_DEBUG_INFO = NO;
238
238
MTL_FAST_MATH = YES;
239
239
SDKROOT = macosx;
···
246
246
isa = XCBuildConfiguration;
247
247
buildSettings = {
248
248
CODE_SIGN_STYLE = Automatic;
249
+
MACOSX_DEPLOYMENT_TARGET = 14.0;
249
250
PRODUCT_NAME = "$(TARGET_NAME)";
250
251
SWIFT_VERSION = 5.0;
251
252
};
···
255
256
isa = XCBuildConfiguration;
256
257
buildSettings = {
257
258
CODE_SIGN_STYLE = Automatic;
259
+
MACOSX_DEPLOYMENT_TARGET = 14.0;
258
260
PRODUCT_NAME = "$(TARGET_NAME)";
259
261
SWIFT_VERSION = 5.0;
260
262
};
+42
-64
OCR/main.swift
+42
-64
OCR/main.swift
···
8
8
import Vision
9
9
import Cocoa
10
10
11
-
var MODE = VNRequestTextRecognitionLevel.accurate // or .fast
12
-
var USE_LANG_CORRECTION = false
13
-
var REVISION:Int
14
-
if #available(macOS 11, *) {
15
-
REVISION = VNRecognizeTextRequestRevision2
16
-
} else {
17
-
REVISION = VNRecognizeTextRequestRevision1
18
-
}
19
-
11
+
var MODE = VNRequestTextRecognitionLevel.accurate
12
+
var USE_LANG_CORRECTION = true
13
+
var REVISION = VNRecognizeTextRequestRevision3
20
14
func main(args: [String]) -> Int32 {
21
-
22
-
if CommandLine.arguments.count == 2 {
23
-
if args[1] == "--langs" {
24
-
let request = VNRecognizeTextRequest.init()
25
-
request.revision = REVISION
26
-
request.recognitionLevel = VNRequestTextRecognitionLevel.accurate
27
-
let langs = try? request.supportedRecognitionLanguages()
28
-
for lang in langs! {
29
-
print(lang)
30
-
}
31
-
}
32
-
return 0
33
-
}else if CommandLine.arguments.count == 6 {
34
-
let (language, fastmode, languageCorrection, src, dst) = (args[1], args[2],args[3],args[4],args[5])
35
-
let substrings = language.split(separator: ",")
36
-
var languages:[String] = []
37
-
for substring in substrings {
38
-
languages.append(String(substring))
39
-
}
40
-
if fastmode == "true" {
41
-
MODE = VNRequestTextRecognitionLevel.fast
42
-
}else{
43
-
MODE = VNRequestTextRecognitionLevel.accurate
44
-
}
45
-
46
-
if languageCorrection == "true" {
47
-
USE_LANG_CORRECTION = true
48
-
}else{
49
-
USE_LANG_CORRECTION = false
15
+
let language = "en-US"
16
+
var languages:[String] = []
17
+
languages.append(language)
18
+
let url = URL(fileURLWithPath: args[1])
19
+
var files = [URL]()
20
+
if let enumerator = FileManager.default.enumerator(at: url, includingPropertiesForKeys: [.isRegularFileKey], options: [.skipsHiddenFiles, .skipsPackageDescendants]) {
21
+
for case let fileURL as URL in enumerator {
22
+
do {
23
+
let fileAttributes = try fileURL.resourceValues(forKeys:[.isRegularFileKey])
24
+
if fileAttributes.isRegularFile! && fileURL.pathExtension.lowercased().range(of: "(jpe?g|png|heic)$", options: .regularExpression) != nil {
25
+
files.append(fileURL)
26
+
}
27
+
} catch { print(error, fileURL) }
50
28
}
51
-
52
-
guard let img = NSImage(byReferencingFile: src) else {
53
-
fputs("Error: failed to load image '\(src)'\n", stderr)
29
+
// print(files)
30
+
}
31
+
32
+
for file in files {
33
+
guard let img = NSImage(byReferencing: file) as NSImage? else {
34
+
fputs("Error: failed to load image '\(file)'\n", stderr)
54
35
return 1
55
36
}
56
37
guard let imgRef = img.cgImage(forProposedRect: nil, context: nil, hints: nil) else {
57
-
fputs("Error: failed to convert NSImage to CGImage for '\(src)'\n", stderr)
38
+
fputs("Error: failed to convert NSImage to CGImage for '\(file)'\n", stderr)
58
39
return 1
59
40
}
60
-
61
-
41
+
62
42
let request = VNRecognizeTextRequest { (request, error) in
63
43
let observations = request.results as? [VNRecognizedTextObservation] ?? []
64
44
var dict:[String:Any] = [:]
···
81
61
let rect = VNImageRectForNormalizedRect(boundingBox,
82
62
Int(imgRef.width),
83
63
Int(imgRef.height))
84
-
64
+
85
65
line["text"] = string ?? ""
86
66
line["confidence"] = confidence ?? ""
87
67
line["x"] = Int(rect.minX)
···
92
72
allText = allText + (string ?? "")
93
73
index = index + 1
94
74
if index != observations.count {
95
-
allText = allText + "\n"
75
+
allText = allText + "\n"
96
76
}
97
77
}
98
78
dict["lines"] = lines
99
79
dict["text"] = allText
100
80
let data = try? JSONSerialization.data(withJSONObject: dict, options: [])
101
-
let jsonString = String(data: data!,
102
-
encoding: .utf8) ?? "[]"
103
-
try? jsonString.write(to: URL(fileURLWithPath: dst), atomically: true, encoding: String.Encoding.utf8)
81
+
let jsonString = String(data: data!, encoding: .utf8) ?? "[]"
82
+
let dstFile = file.absoluteString + ".json"
83
+
do {
84
+
try jsonString.write(to: URL(string: dstFile)!, atomically: true, encoding: String.Encoding.utf8)
85
+
} catch {
86
+
print("Unexpected error writing JSON: \(error).")
87
+
exit(1)
88
+
}
104
89
}
105
90
request.recognitionLevel = MODE
106
91
request.usesLanguageCorrection = USE_LANG_CORRECTION
···
108
93
request.recognitionLanguages = languages
109
94
//request.minimumTextHeight = 0
110
95
//request.customWords = [String]
111
-
try? VNImageRequestHandler(cgImage: imgRef, options: [:]).perform([request])
112
-
113
-
return 0
114
-
}else{
115
-
print("""
116
-
usage:
117
-
language fastmode languageCorrection image_path output_path
118
-
--langs: list suppported languages
119
-
120
-
example:
121
-
macOCR en false true ./image.jpg out.json
122
-
""")
123
-
return 1
96
+
do {
97
+
try VNImageRequestHandler(cgImage: imgRef, options: [:]).perform([request])
98
+
} catch {
99
+
print("Unexpected error with OCR: \(error).")
100
+
exit(1)
101
+
}
102
+
print("\(file) done")
124
103
}
104
+
return 0
125
105
}
126
-
127
-
128
106
129
107
exit(main(args: CommandLine.arguments))