-
Notifications
You must be signed in to change notification settings - Fork 5
/
FillModelWithData.java
280 lines (253 loc) · 10.5 KB
/
FillModelWithData.java
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
package org.genericsystem.cv.comparator;
import java.io.File;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.util.Arrays;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.function.Function;
import java.util.stream.Stream;
import org.genericsystem.common.Generic;
import org.genericsystem.cv.Img;
import org.genericsystem.cv.Zones;
import org.genericsystem.cv.model.Doc;
import org.genericsystem.cv.model.Doc.DocInstance;
import org.genericsystem.cv.model.DocClass;
import org.genericsystem.cv.model.DocClass.DocClassInstance;
import org.genericsystem.cv.model.ImgFilter.ImgFilterInstance;
import org.genericsystem.cv.model.MeanLevenshtein;
import org.genericsystem.cv.model.ImgFilter;
import org.genericsystem.cv.model.Score;
import org.genericsystem.cv.model.Score.ScoreInstance;
import org.genericsystem.cv.model.ZoneGeneric;
import org.genericsystem.cv.model.ZoneGeneric.ZoneInstance;
import org.genericsystem.cv.model.ZoneText;
import org.genericsystem.cv.model.ZoneText.ZoneTextInstance;
import org.genericsystem.kernel.Engine;
import org.opencv.core.Core;
import org.opencv.core.Scalar;
import org.opencv.imgcodecs.Imgcodecs;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
/**
* The FillModelWithData class can analyze an image (or a batch of images) and
* store all the OCR text for each zone and each document in GS.
*
* @author Pierrik Lassalas
*
*/
public class FillModelWithData {
private static Logger log = LoggerFactory.getLogger(FillModelWithData.class);
private static final String gsPath = System.getenv("HOME") + "/genericsystem/gs-cv_model3/";
private final static Engine engine = new Engine(gsPath, Doc.class, ImgFilter.class, ZoneGeneric.class,
ZoneText.class, Score.class, MeanLevenshtein.class);
private static final String docType = "id-fr-front";
static {
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
System.out.println("OpenCV core library loaded");
}
public static void main(String[] mainArgs) {
engine.newCache().start();
compute();
// cleanModel();
engine.close();
// doImgOcr(Paths.get("/home/middleware/git/genericsystem2015/gs-cv/classes/id-fr-front"));
}
/**
* This Map will contain the names of the filter that will be applied to a
* specified {@link Img}.
*
* @return - a Map containing the filter names as key, and a
* {@link Function} that will apply the specified algorithm to an
* Img.
*/
public static Map<String, Function<Img, Img>> getFiltersMap() {
final Map<String, Function<Img, Img>> map = new HashMap<>();
map.put("original", Img::bgr2Gray);
map.put("reality", Img::bgr2Gray);
// map.put("bernsen", Img::bernsen);
// map.put("equalizeHisto", Img::equalizeHisto);
// map.put("equalizeHistoAdaptative", Img::equalizeHistoAdaptative);
return map;
}
public static void doImgOcr(Path imagePath) {
final Path imgClassDirectory = imagePath.getParent();
final String docType = imgClassDirectory.getName(imgClassDirectory.getNameCount() - 1).toString();
final String imgDirectory = imgClassDirectory.toString() + "/ref2/";
log.info("imgDirectory = {} ", imgDirectory);
engine.newCache().start();
// Find and save the doc class
DocClass docClass = engine.find(DocClass.class);
DocClassInstance docClassInstance = docClass.addDocClass(docType);
engine.getCurrentCache().flush();
// Get the filters and the predefined zones
final Map<String, Function<Img, Img>> imgFilters = getFiltersMap();
final Zones zones = Zones.load(imgClassDirectory.toString());
// Process the image file
File file = new File(imagePath.toString());
initComputation(imgClassDirectory.toString(), docType, imgFilters, zones);
processFile(file, docClassInstance, zones, imgFilters.entrySet().stream());
engine.getCurrentCache().flush();
engine.close();
}
private static void compute() {
final String imgClassDirectory = "classes/" + docType;
final String imgDirectory = imgClassDirectory + "/ref2/";
log.info("imgClassDirectory = {} ", imgClassDirectory);
// Save the current document class
DocClass docClass = engine.find(DocClass.class);
DocClassInstance docClassInstance = docClass.addDocClass(docType);
// Set all the filter names
Map<String, Function<Img, Img>> imgFilters = filterOptimizationMap();
// Load the accurate zones
final Zones zones = Zones.load(imgClassDirectory);
initComputation(imgClassDirectory, docType, imgFilters, zones);
// Process each file in folder imgDirectory
Arrays.asList(new File(imgDirectory).listFiles((dir, name) -> name.endsWith(".png"))).forEach(file -> {
processFile(file, docClassInstance, zones, imgFilters.entrySet().stream());
engine.getCurrentCache().flush();
});
engine.getCurrentCache().flush();
}
private static void initComputation(String imgClassDirectory, String docType,
Map<String, Function<Img, Img>> imgFilters, Zones zones) {
DocClass docClass = engine.find(DocClass.class);
ImgFilter imgFilter = engine.find(ImgFilter.class);
DocClassInstance docClassInstance = docClass.getDocClass(docType);
// Save the zones
zones.getZones().forEach(z -> {
log.info("Adding zone n° {}", z.getNum());
docClassInstance.addZone(z.getNum(), z.getRect().x, z.getRect().y, z.getRect().width, z.getRect().height);
});
// Save the filternames
imgFilters.entrySet().forEach(entry -> {
log.info("Adding algorithm : {} ", entry.getKey());
imgFilter.addImgFilter(entry.getKey());
});
// Persist the changes
engine.getCurrentCache().flush();
}
private static void processFile(File file, DocClassInstance docClassInstance, Zones zones,
Stream<Entry<String, Function<Img, Img>>> imgFilters) {
log.info("\nProcessing file: {}", file.getName());
Generic doc = engine.find(Doc.class);
ZoneText zoneText = engine.find(ZoneText.class);
ImgFilter imgFilter = engine.find(ImgFilter.class);
// Save the current file (document)
DocInstance docInstance = docClassInstance.addDoc(doc, file.getName());
engine.getCurrentCache().flush();
// Draw the image's zones + numbers
Img imgCopy = new Img(Imgcodecs.imread(file.getPath()));
zones.draw(imgCopy, new Scalar(0, 255, 0), 3);
zones.writeNum(imgCopy, new Scalar(0, 0, 255), 3);
// Copy the images to the resources folder
// TODO implement a filter mechanism to avoid creating
// duplicates in a public folder
Imgcodecs.imwrite(System.getProperty("user.dir") + "/src/main/resources/" + file.getName(),
imgCopy.getSrc());
// Create a map of Imgs
Img originalImg = new Img(Imgcodecs.imread(file.getPath()));
Map<String, Img> imgs = new HashMap<>();
imgFilters.forEach(entry -> {
log.info("Applying algorithm {}...", entry.getKey());
Img img = null;
if ("original".equals(entry.getKey()) || "reality".equals(entry.getKey()))
img = originalImg.bgr2Gray();
else
img = entry.getValue().apply(originalImg);
if (null != img)
imgs.put(entry.getKey(), img);
else
log.error("An error as occured for image {} and filter {}", file.getName(), entry.getKey());
});
// Process each zone
zones.getZones().forEach(z -> {
log.info("Zone n° {}", z.getNum());
ZoneInstance zoneInstance = docClassInstance.getZone(z.getNum());
imgs.entrySet().forEach(entry -> {
if ("reality".equals(entry.getKey()) && null != zoneText.getZoneText(docInstance, zoneInstance,
imgFilter.getImgFilter(entry.getKey()))) {
// Do not proceed to OCR if the real values are known
} else {
String ocrText = z.ocr(entry.getValue());
zoneText.addZoneText(ocrText.trim(), docInstance, zoneInstance,
imgFilter.getImgFilter(entry.getKey()));
}
});
engine.getCurrentCache().flush();
// Call the garbage collector to free the resources used by
// OpenCV
System.gc();
});
}
private static Map<String, Function<Img, Img>> filterOptimizationMap() {
final Map<String, Function<Img, Img>> imgFilters = new HashMap<>();
// Niblack
// List<Integer> blockSizes = Arrays.asList(new Integer[] { 7, 9, 11,
// 15, 17, 21, 27, 37 });
// List<Double> ks = Arrays.asList(new Double[] { -1.0, -0.8, -0.6,
// -0.5, -0.4, -0.3, -0.2, -0.1, 0.0, 0.1 });
// Sauvola, Nick
List<Integer> blockSizes = Arrays.asList(new Integer[] { 7, 9, 11, 15, 17, 21, 27, 37 });
List<Double> ks = Arrays.asList(new Double[] { 0.0, 0.1, 0.2, 0.3, 0.4 });
// Wolf
// List<Integer> blockSizes = Arrays.asList(new Integer[] { 7, 9, 11,
// 15, 17, 21, 27, 37 });
// List<Double> ks = Arrays.asList(new Double[] { -0.25, -0.2, -0.15,
// 0.1, -0.05, 0.0 });
for (Integer bs : blockSizes) {
for (Double k : ks) {
imgFilters.put("nick" + "_" + bs + "_" + k.toString().replace("-", "m"),
img -> img.niblackThreshold(bs, k));
}
}
imgFilters.put("reality", Img::bgr2Gray);
imgFilters.put("original", Img::bgr2Gray);
return imgFilters;
}
@SuppressWarnings({ "unused", "unchecked", "rawtypes" })
private static void cleanModel() {
// TODO: not working as expected
// TODO: extract the "reality" infos, then erase the model and rebuild?
System.out.println("Cleaning model...");
final String imgClassDirectory = "classes/" + docType;
final String imgDirectory = imgClassDirectory + "/ref2/";
// Get the necessary classes from the engine
DocClass docClass = engine.find(DocClass.class);
Generic doc = engine.find(Doc.class);
ZoneText zoneText = engine.find(ZoneText.class);
ImgFilter imgFilter = engine.find(ImgFilter.class);
Score score = engine.find(Score.class);
MeanLevenshtein ml = engine.find(MeanLevenshtein.class);
// Save the current document class
Generic currentDocClass = engine.find(DocClass.class).getInstance(docType);
List<ImgFilterInstance> imgFilterInstances = (List) imgFilter.getInstances()
.filter(f -> !"reality".equals(f.getValue())).toList();
List<ZoneInstance> zoneInstances = (List) currentDocClass.getHolders(engine.find(ZoneGeneric.class)).toList();
List<DocInstance> docInstances = (List) currentDocClass.getHolders(engine.find(Doc.class)).toList();
docInstances.forEach(currentDoc -> {
imgFilterInstances.forEach(i -> {
zoneInstances.forEach(z -> {
ZoneTextInstance zti = zoneText.getZoneText(currentDoc, z, i);
if (zti != null)
zti.remove();
});
engine.getCurrentCache().flush();
});
});
imgFilterInstances.forEach(i -> {
zoneInstances.forEach(z -> {
ScoreInstance scoreInst = score.getScore(z, i);
if (scoreInst != null) {
scoreInst.getHolder(ml).remove();
scoreInst.remove();
}
});
i.remove();
});
engine.getCurrentCache().flush();
System.out.println("Done!");
}
}