OCR Recognition

1
function main() {
2
 let appleVision = {"type": "appleVision", "level": "accurate", "languages": "zh-Hans,en-US"}
3
 let inited = ocr.initOcr(appleVision)
4
 logd("Init result -" + inited);
5
 if (!inited) {
6
 loge("error : " + ocr.getErrorMsg());
7
 return;
8
 }
9
 for (var ix = 0; ix < 20; ix++) {
10
 // Read a bitmap
11
 let img = image.captureFullScreen();
12
 if (img == null || img == undefined || img.uuid == null || img.uuid == undefined || img.uuid == "") {
13
 loge("Failed to read image");
14
 continue;
15
 }
16
 console.time("1")
17
 logd("start---ocr");
18
 // Recognize the image
19
 let result = ocr.ocrImage(img, 20 * 1000, {});
20
 logd(result)
21
 if (result) {
22
 logd("OCR result -> " + JSON.stringify(result));
23
 for (var i = 0; i < result.length; i++) {
24
 var value = result[i];
25
 logd("Text : " + value.label + " x: " + value.x + " y: " + value.y + " width: " + value.width + " height: " + value.height);
26
 }
27
 } else {
28
 logw("No result recognized");
29
 }
30

31
 logd("Elapsed: " + console.timeEnd(1) + " ms")
32
 image.recycle(img)
33
 sleep(1000);
34
 logd("ix = " + ix)
35
 }
36
 // Release all resources
37
 ocr.releaseAll();
38
}
39

40
main();

1
 [
2
 {
3
 "label": "奇趣装扮三阶盘化",
4
 "confidence": 0.48334712,
5
 "x": 11,
6
 "y": 25,
7
 "width": 100,
8
 "height": 100
9
 }
10
]

1
See common code examples
2
OCR initialization

1
See common code examples
2
OCR initialization

1
See common code examples
2
OCR initialization

1
// See code examples

1
 {
2
 "type": "ocrLite",
3
 "padding": 10,
4
 "maxSideLen": 0,
5
 "boxScoreThresh": 0.6,
6
 "boxThresh": 0.3,
7
 "unClipRatio": 1.6,
8
 "doAngle": 0,
9
 "mostAngle": 0
10
}

1
* numThread: thread count — `-1` = all, `-2` = half of device CPUs, `0` = not set (program decides); `-2` recommended<br/>
2
* padding: white border around image to improve recognition; increase when text boxes do not fully enclose text. Default 50.<br/>
3
* maxSideLen: scale by longest image side; larger = slower but more accurate, smaller = faster but less accurate; `0` = no scaling.<br/>
4
* boxScoreThresh: text box confidence threshold; decrease when text boxes do not fully enclose text<br/>
5
* boxThresh: same as above; tune experimentally.<br/>
6
* unClipRatio: single text box size multiplier; larger values produce bigger boxes.<br/>
7
* doAngle: enable (1) / disable (0) text orientation detection; needed only for upside-down images (rotated 90°–270°). Default off.<br/>
8
* mostAngle: enable (1) / disable (0) angle voting; has no effect when orientation detection is disabled. Default off.<br/>

1
 {
2
 "type": "paddleOcrOnline",
3
 "ocrType": "ONNX_PPOCR_V3",
4
 "padding": 50,
5
 "maxSideLen": 0,
6
 "boxScoreThresh": 0.5,
7
 "boxThresh": 0.3,
8
 "unClipRatio": 1.6,
9
 "doAngleFlag": 0,
10
 "mostAngleFlag": 0
11
}

1
* ocrType: model — ONNX_PPOCR_V3, ONNX_PPOCR_V4, NCNN_PPOCR_V3
2
* serverUrl: Paddle OCR server address; can deploy on another PC and connect from control center, e.g. 192.168.2.8:9022; change IP when deployed on PC; port 9022 is optional
3
* padding: white border around image to improve recognition; increase when text boxes do not fully enclose text. Default 50.<br/>
4
* maxSideLen: scale by longest image side; larger = slower but more accurate, smaller = faster but less accurate; `0` = no scaling.<br/>
5
* boxScoreThresh: text box confidence threshold; decrease when text boxes do not fully enclose text<br/>
6
* boxThresh: same as above; tune experimentally.<br/>
7
* unClipRatio: single text box size multiplier; larger values produce bigger boxes.<br/>
8
* doAngleFlag: enable (1) / disable (0) text orientation detection; needed only for upside-down images (rotated 90°–270°). Default off.<br/>
9
* mostAngleFlag: enable (1) / disable (0) angle voting; has no effect when orientation detection is disabled. Default off.<br/>
10
* limit: OCR requests per second; default 1000. Lower to reduce CPU usage<br/>
11
* checkImage: verify data is an image (1 yes, 0 no); default off.<br/>

1
function main() {
2
 // Release all OCR resources at start to avoid leaks from previous runs
3
 ocrMut.releaseAll();
4
 logd("Start script...")
5

6
 // Initialize an instance
7
 let ocrtest = ocrMut.newOcr();
8
 let vision = {"type": "appleVision", "level": "accurate", "languages": "zh-Hans,en-US"}
9
 // paddleOcr parameters
10
 let paddleOcrOnline = {
11
 "type": "paddleOcrOnline",
12
 "ocrType": "ONNX_PPOCR_V3",
13
 "serverUrl": "192.168.2.13:9022",
14
 "limit": 12,
15
 "checkImage": "1",
16
 "padding": 200
17
 }
18
 let ocrLite = {"type": "ocrLite","numThread":2}
19
 let paddleLiteOcrMap = {"type": "paddleLiteOcr","cpuThreadNum":2,"cpuPowerMode":"LITE_POWER_FULL"}
20
 let inited = ocrtest.initOcr(ocrLite)
21
 // let inited = ocrtest.initOcr(paddleLiteOcrMap)
22
 if (!inited) {
23
 loge("inited ocr error : " + ocrtest.getErrorMsg())
24
 return
25
 } else {
26
 logd("ocr inited ok")
27
 }
28
 for (let i = 0; i < 3; i++) {
29
 let img = image.captureFullScreen()
30
 let ocrResult = ocrtest.ocrImage(img, 20000, null)
31
 logd("ocrResult " + JSON.stringify(ocrResult));
32
 if (ocrResult) {
33
 logd("OCR result -> " + JSON.stringify(ocrResult));
34
 for (var j = 0; j < ocrResult.length; j++) {
35
 var value = ocrResult[j];
36
 logd("Text : " + value.label + " x: " + value.x + " y: " + value.y + " width: " + value.width + " height: " + value.height);
37
 }
38
 } else {
39
 logw("No result recognized");
40
 }
41
 image.recycle(img)
42

43
 sleep(2000)
44
 }
45
 // Release when script finishes; no need to release after every use
46
 ocrtest.releaseAll()
47
}
48

49
main();

1
function main() {
2
 // Release all OCR resources at start to avoid leaks from previous runs
3
 ocrMut.releaseAll();
4
 logd("Start")
5
 let ts = file.getSandBoxDir()
6
 let tessdataDir = ts + "/tessdata"
7
 logd("tessdataDir=> ", tessdataDir)
8
 file.mkdirs(tessdataDir)
9
 // Save traineddata datasets to res folder; auto-copied to tessdataDir
10
 // Copy English dataset
11
 let saved = saveResToFile("eng.traineddata", tessdataDir + "/eng.traineddata")
12
 // Copy Chinese dataset
13
 let saved2 = saveResToFile("chi_sim.traineddata", tessdataDir + "/chi_sim.traineddata")
14
 logd("saved ", saved)
15
 if (!saved) {
16
 logd("copy eng error")
17
 return;
18
 }
19
 if (!saved2) {
20
 logd("copy chi_sim error")
21
 return;
22
 }
23
 let tessocr = ocrMut.newOcr()
24
 // Initialize Chinese + English
25
 let intx = tessocr.initOcr({
26
 "type": "ocrLite",
27
 "path": tessdataDir,
28
 "language": "eng+chi_sim",
29
 "rilLevel": 2,
30
 })
31

32
 logd("tessocr initOcr " + intx)
33
 if (!intx) {
34
 logd(tessocr.getErrorMsg());
35
 return
36
 }
37

38
 for (let i = 0; i < 10; i++) {
39
 // Can use screenshot here instead
40
 //let aa = image.captureFullScreen();
41
 let aa = readResAutoImage("2.png")
42

43
 console.time(1)
44
 let rse = tessocr.ocrImage(aa, 10 * 1000, {})
45
 logd("Elapsed time - ", console.timeEnd(1))
46
 image.recycle(aa)
47

48
 logd(JSON.stringify(rse));
49
 for (let i = 0; i < rse.length; i++) {
50
 let a = rse[i]
51
 logd(JSON.stringify(a))
52
 let b = a.x + "," + a.y + "," + (a.x + a.width) + "," + (a.y + a.height)
53
 logd(a.label, b)
54
 }
55

56
 }
57

58
 tessocr.releaseAll()
59

60
 logd("end--")
61
}
62

63

64
main();

1
function main() {
2
 // Release all OCR resources at start to avoid leaks from previous runs
3
 ocrMut.releaseAll();
4
 logd("Start script...")
5
 let ocrtest = ocrMut.newOcr()
6
 let modelPath = file.getSandBoxFilePath("")
7
 let labelPath = file.getSandBoxFilePath("ppocrv5_mobile_labels.txt")
8
 let detModelFilename = "ch_PP-OCRv5_mobile_det.onnx"
9
 let recModelFilename = "ch_PP-OCRv5_rec_mobile_infer.onnx"
10
 let clsModelFilename = "ch_ppocr_mobile_v2.0_cls_infer.onnx"
11
 // Built-in OCR model
12
 let paddleOnnxOcrMap1 = {"type": "paddleOnnxOcr", "cpuThreadNum": 2, "padding": 10, "maxSideLen": 960}
13

14
 // External model
15
 let paddleOnnxOcrMap2 = {
16
 "type": "paddleOnnxOcr", "cpuThreadNum": 2, "padding": 10, "maxSideLen": 960,
17
 "modelPath": modelPath,
18
 "labelPath": labelPath,
19
 "detModelFilename": detModelFilename,
20
 "recModelFilename": recModelFilename,
21
 "clsModelFilename": clsModelFilename,
22
 }
23
 let inited = ocrtest.initOcr(paddleOnnxOcrMap1)
24
 if (!inited) {
25
 loge("inited ocr error : " + ocrtest.getErrorMsg())
26
 return
27
 } else {
28
 logd("ocr inited ok")
29
 }
30
 for (let i = 0; i < 100; i++) {
31
 let img = image.captureFullScreen()
32

33

34
 console.time(1)
35

36
 let ocrResult = ocrtest.ocrImage(img, 20000, {"cpuThreadNum": 2, "padding": 10, "maxSideLen": 960})
37

38
 logd("ocrResult " + JSON.stringify(ocrResult));
39
 if (ocrResult) {
40
 logd("OCR result -> " + JSON.stringify(ocrResult));
41
 for (var j = 0; j < ocrResult.length; j++) {
42
 var value = ocrResult[j];
43
 logd("Text : " + value.label + " " + value.x + "," + value.y + "," + (value.x + value.width) + "," + (value.y + value.height));
44
 }
45
 } else {
46
 logw("No result recognized");
47
 }
48
 logd("Elapsed: {} ms", console.timeEnd(1))
49
 sleep(100)
50
 image.recycle(img)
51
 sleep(200)
52
 }
53
 // Release when script finishes; no need to release after every use
54
 ocrtest.releaseAll()
55
}
56

57

58
main()

1
function main() {
2
 // Release all OCR resources at start to avoid leaks from previous runs
3
 ocrMut.releaseAll();
4
 logd("Start script...")
5
 let ocrtest = ocrMut.newOcr()
6
 let modelPath = file.getSandBoxFilePath("")
7
 let labelPath = file.getSandBoxFilePath("ppocrv6_small_labels.txt")
8
 let detModelFilename = "PP-OCRv6_small_det.onnx"
9
 let recModelFilename = "PP-OCRv6_small_rec.onnx"
10
 // v6 angle classification reuses v5 cls
11
 let clsModelFilename = "ch_ppocr_mobile_v2.0_cls_infer.onnx"
12

13
 // Fast mode (recommended for daily UI screenshots): smaller longest side, angle detection off, more CPU
14
 let paddleOnnxOcrV6MapFast = {
15
 "type": "paddleOnnxOcrV6",
16
 "cpuThreadNum": -2,
17
 "padding": 10,
18
 "maxSideLen": 640,
19
 "doAngle": 0,
20
 "mostAngle": 0
21
 }
22

23
 // Default accuracy-oriented: maxSideLen 960, orientation detection enabled
24
 let paddleOnnxOcrV6Map1 = {"type": "paddleOnnxOcrV6", "cpuThreadNum": 2, "padding": 10, "maxSideLen": 960}
25

26
 // External model (place onnx / dictionary in sandbox yourself)
27
 let paddleOnnxOcrV6Map2 = {
28
 "type": "paddleOnnxOcrV6", "cpuThreadNum": 2, "padding": 10, "maxSideLen": 960,
29
 "modelPath": modelPath,
30
 "labelPath": labelPath,
31
 "detModelFilename": detModelFilename,
32
 "recModelFilename": recModelFilename,
33
 "clsModelFilename": clsModelFilename,
34
 }
35
 let inited = ocrtest.initOcr(paddleOnnxOcrV6MapFast)
36
 if (!inited) {
37
 loge("inited ocr error : " + ocrtest.getErrorMsg())
38
 return
39
 } else {
40
 logd("ocr v6 inited ok")
41
 }
42
 for (let i = 0; i < 100; i++) {
43
 let img = image.captureFullScreen()
44

45
 console.time(1)
46

47
 let ocrResult = ocrtest.ocrImage(img, 20000, {"cpuThreadNum": -2, "padding": 10, "maxSideLen": 640, "doAngle": 0, "mostAngle": 0})
48

49
 logd("ocrResult " + JSON.stringify(ocrResult));
50
 if (ocrResult) {
51
 logd("OCR result -> " + JSON.stringify(ocrResult));
52
 for (var j = 0; j < ocrResult.length; j++) {
53
 var value = ocrResult[j];
54
 logd("Text : " + value.label + " " + value.x + "," + value.y + "," + (value.x + value.width) + "," + (value.y + value.height));
55
 }
56
 } else {
57
 logw("No result recognized");
58
 }
59
 logd("Elapsed: {} ms", console.timeEnd(1))
60
 sleep(100)
61
 image.recycle(img)
62
 sleep(200)
63
 }
64
 // Release when script finishes; no need to release after every use
65
 ocrtest.releaseAll()
66
}
67

68

69
main()

1
function main() {
2
 // Release all OCR resources at start to avoid leaks from previous runs
3
 ocrMut.releaseAll();
4
 logd("Start script...")
5
 let ocrtest = ocrMut.newOcr()
6
 let modelPath = file.getSandBoxFilePath("")
7
 let detName = "det"
8
 let recName = "rec"
9
 // Built-in OCR model
10
 let paddleNcnnOcrMap1 = {"type": "paddleNcnnOcrV5", "numThread": 2, "padding": 32, "maxSideLen": 640}
11

12
 // External model
13
 let paddleOnnxOcrMap2 = {
14
 "type": "paddleNcnnOcrV5", "numThread": 2, "padding": 32, "maxSideLen": 640,
15
 "modelsDir": modelPath,
16
 "keysName": "keys.txt",
17
 "detName": detName,
18
 "recName": recName
19
 }
20
 let inited = ocrtest.initOcr(paddleNcnnOcrMap1)
21
 if (!inited) {
22
 loge("inited ocr error : " + ocrtest.getErrorMsg())
23
 return
24
 } else {
25
 logd("ocr inited ok")
26
 }
27
 for (let i = 0; i < 100; i++) {
28
 let img = image.captureFullScreen()
29

30

31
 console.time(1)
32

33
 // Dynamic parameters can also be set here
34
 let ocrResult = ocrtest.ocrImage(img, 20000, {"numThread":2,"padding":32})
35

36
 logd("ocrResult " + JSON.stringify(ocrResult));
37
 if (ocrResult) {
38
 logd("OCR result -> " + JSON.stringify(ocrResult));
39
 for (var j = 0; j < ocrResult.length; j++) {
40
 var value = ocrResult[j];
41
 logd("Text : " + value.label + " " + value.x + "," + value.y + "," + (value.x + value.width) + "," + (value.y + value.height));
42
 }
43
 } else {
44
 logw("No result recognized");
45
 }
46
 logd("Elapsed: {} ms", console.timeEnd(1))
47
 sleep(100)
48
 image.recycle(img)
49
 sleep(200)
50
 }
51
 // Release when script finishes; no need to release after every use
52
 ocrtest.releaseAll()
53
}
54

55

56
main()

1
 [
2
 {
3
 "label": "奇趣装扮三阶盘化",
4
 "confidence": 0.48334712,
5
 "x": 11,
6
 "y": 25,
7
 "width": 100,
8
 "height": 100
9
 }
10
]

1
See common code examples
2
OCR initialization

1
See common code examples
2
OCR initialization

1
See common code examples
2
OCR initialization

1
function httpPaddleOcr(filePath) {
2
 // OCR service address
3
 // See parameter descriptions above for other options
4
 let url = "http://192.168.2.13:9022/devapi/uploadOcr"
5
 let ocrType = "ONNX_PPOCR_V3"
6
 let limit = "1000"
7
 let ocrParam = {
8
 "padding": 50,
9
 "maxSideLen": 0,
10
 "boxScoreThresh": 0.5,
11
 "boxThresh": 0.3,
12
 "unClipRatio": 1.6,
13
 "doAngleFlag": 0,
14
 "mostAngleFlag": 0
15
 }
16
 ocrParam = utils.base64Encode(JSON.stringify(ocrParam));
17
 let param = {
18
 "ocrType": ocrType,
19
 "limit": limit,
20
 "ocrParam": ocrParam
21
 };
22
 let files = {
23
 "file": filePath
24
 }
25
 let result = http.httpPost(url, param, files, 20 * 1000, {"User-Agent": "test"});
26
 if (result == null || result == undefined || result == "") {
27
 return null;
28
 }
29

30
 try {
31
 result = JSON.parse(result)
32
 return result["data"]
33
 } catch (e) {
34
 return null;
35
 }
36

37
}
38

39
function callPaddleOcrTest() {
40
 let img = image.captureFullScreen()
41
 if (!img) {
42
 loge("Screenshot failed");
43
 return
44
 }
45
 let filePath = file.getSandBoxFilePath("ocrtmp.jpg")
46
 image.saveTo(img, filePath)
47
 let result = httpPaddleOcr(filePath);
48
 logd("result " + JSON.stringify(result));
49

50

51
}
52

53
callPaddleOcrTest()