File size: 10,533 Bytes
3f42a6f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import cv2, string, os, time
import numpy as np
from edocr2 import tools
from pdf2image import convert_from_path
           
def ocr_drawing(file_path, recognizer_gdt, dimension_tuple, #Must have
                autoframe = True, language = 'eng', binary_thres= 127, frame_thres = 0.85, #general
                GDT_thres = 0.02, #GD&Ts
                cluster_thres= 15, max_char = 15, max_img_size=2048, #Dimensions
                output_path='.', save_mask=False, save_raw_output=False, backg_save=False #Output
                ):
    #Read file
    if file_path.endswith('.pdf') or file_path.endswith(".PDF"):
        img = convert_from_path(file_path)
        img = np.array(img[0])
    else:
        img = cv2.imread(file_path)

    #Layer Segmentation
    times = []
    start_time = time.time()
    _, frame, gdt_boxes, tables, dim_boxes = tools.layer_segm.segment_img(img, autoframe = autoframe, frame_thres=frame_thres, GDT_thres = GDT_thres, binary_thres= binary_thres)
    end_time = time.time()
    times.append(end_time-start_time)
    print('Segmentation Done')
    #OCR Tables
    process_img = img.copy()
    table_results, updated_tables, process_img = tools.ocr_pipelines.ocr_tables(tables, process_img, language=language)
    end_time = time.time()
    times.append(end_time-sum(times)-start_time)
    print('Prediction on Tables Done')
    #GD&T OCR
    gdt_results, updated_gdt_boxes, process_img = tools.ocr_pipelines.ocr_gdt(process_img, gdt_boxes,recognizer=recognizer_gdt)
    end_time = time.time()
    times.append(end_time-sum(times)-start_time)
    print('Prediction on GD&T Done')
    #Dimension  OCR
    if frame:
        process_img = process_img[frame.y : frame.y + frame.h, frame.x : frame.x + frame.w]
    
    dimensions, other_info, process_img, _ = tools.ocr_pipelines.ocr_dimensions(process_img, dimension_tuple[0], dimension_tuple[1], dimension_tuple[2],
                                                                                frame, dim_boxes, max_img_size=max_img_size, cluster_thres=cluster_thres, backg_save=backg_save)
    end_time = time.time()
    times.append(end_time-sum(times)-start_time)
    print('Prediction on dimensions and extra information Done')
    #Saving
    if save_mask or save_raw_output:
        filename = os.path.splitext(os.path.basename(file_path))[0]
        output_path = os.path.join(output_path, filename)
        os.makedirs(output_path, exist_ok=True)

    if save_mask:
        mask_img = tools.output_tools.mask_img(img, updated_gdt_boxes, updated_tables, dimensions, frame, other_info)
        cv2.imwrite(os.path.join(output_path, filename + '_mask.png'), mask_img)
        print('Mask saved')

    table_results, gdt_results, dimensions, other_info  = tools.output_tools.process_raw_output(output_path, table_results, gdt_results, dimensions, other_info, save=save_raw_output)
    end_time = time.time()
    times.append(end_time-sum(times)-start_time)
    print('Raw output saved')

    return {'tab': table_results, 'gdts': gdt_results, 'dim': dimensions, 'other': other_info}, times, updated_tables, img, process_img

def ocr_one_drawing(file_path = '/home/javvi51/edocr2/tests/test_samples/4132864.jpg'):

    #Session Loading
    start_time = time.time()
    from edocr2.keras_ocr.recognition import Recognizer
    from edocr2.keras_ocr.detection import Detector

    gdt_model = 'edocr2/models/recognizer_gdts.keras'
    recognizer_gdt = Recognizer(alphabet=tools.ocr_pipelines.read_alphabet(gdt_model))
    recognizer_gdt.model.load_weights(gdt_model)

    dim_model = 'edocr2/models/recognizer_dimensions_2.keras'
    alphabet_dim = tools.ocr_pipelines.read_alphabet(dim_model)
    recognizer_dim = Recognizer(alphabet=alphabet_dim)
    recognizer_dim.model.load_weights(dim_model)

    detector = Detector()
    #detector.model.load_weights('edocr2/models/detector_8_31.keras')

    #Warming up models:
    dummy_image = np.zeros((1, 1, 3), dtype=np.float32)
    _ = recognizer_gdt.recognize(dummy_image)
    _ = recognizer_dim.recognize(dummy_image)
    dummy_image = np.zeros((32, 32, 3), dtype=np.float32)
    _ = detector.detect([dummy_image])
    end_time = time.time()

    kwargs = {
        #General
        'file_path': file_path, #MUST: Image or pdf path to OCR
        'binary_thres': 127, #Pixel value (0-255) to detect contourns, i.e, identify rectangles in the image
        'language': 'eng', #Language of the drawing, require installation of tesseract speficic language if not english
        'autoframe' : False, #Do we want to spot a frame as the maximum rectangle?
        'frame_thres': 0.95, #Frame boundary in % of img, if autoframe, this setting is overruled
        #GD&T
        'recognizer_gdt': recognizer_gdt, #MUST: A Tuple with (gdt alphabet, model path)
        'GDT_thres': 0.02, #Maximum porcentage of the image area to consider a cluster of rectangles a GD&T box
        #Dimensions
        'dimension_tuple': (detector, recognizer_dim, alphabet_dim), #MUST: A Tuple with (detector, dimension alphabet, model path)
        'cluster_thres': 20, #Minimum distance in pixels between two text predictions to consider the same text box
        'max_char': 15, #Max number of characters to consider a text prediction a dimension, otherwise -> other info
        'max_img_size': 2048, #Max size after applying scale for the img patch, bigger, better prediction and higher times
        #Output
        'backg_save': False, #Option to save the background once all text and boxes have been removed, for synth training purposes
        'output_path': '.', #Output path
        'save_mask': True, #Option to save the mask output
        'save_raw_output': True, #Option to save raw ouput, i.e, OCR text and box position,
        }
    
    results, times, _ = ocr_drawing(**kwargs)
    final_time = time.time()
    print(
    "Session Timing Report:\n"
    "Loading session:           {:.3f} s\n"
    "----------------------\n"
    "Drawing segmentation:      {:.3f} s\n"
    "Table prediction:          {:.3f} s\n"
    "GD&T prediction:           {:.3f} s\n"
    "Dimension & other info:    {:.3f} s\n"
    "Saving & processing:       {:.3f} s\n"
    "----------------------\n"
    "OCR in {}: {:.3f} s\n"
    "----------------------\n"
    "Total time:                {:.3f} s\n"
    .format(end_time - start_time, times[0], times[1], times[2], times[3], times[4], os.path.basename(file_path), sum(times), final_time - start_time)
)

def ocr_folder(folder_path = '/home/javvi51/edocr2/tests/test_samples/Washers'):
       
    #Session Loading
    start_time = time.time()
    from edocr2.keras_ocr.recognition import Recognizer
    from edocr2.keras_ocr.detection import Detector

    gdt_model = 'edocr2/models/recognizer_gdts.keras'
    recognizer_gdt = Recognizer(alphabet=tools.ocr_pipelines.read_alphabet(gdt_model))
    recognizer_gdt.model.load_weights(gdt_model)

    dim_model = 'edocr2/models/recognizer_dimensions_2.keras'
    alphabet_dim = tools.ocr_pipelines.read_alphabet(dim_model)
    recognizer_dim = Recognizer(alphabet=alphabet_dim)
    recognizer_dim.model.load_weights(dim_model)

    detector = Detector()
    #detector.model.load_weights('path/to/custom/detector')

    #Warming up models:
    dummy_image = np.zeros((1, 1, 3), dtype=np.float32)
    _ = recognizer_gdt.recognize(dummy_image)
    _ = recognizer_dim.recognize(dummy_image)
    dummy_image = np.zeros((32, 32, 3), dtype=np.float32)
    _ = detector.detect([dummy_image])
    end_time = time.time()

    file_paths = [os.path.join(folder_path, filename) for filename in os.listdir(folder_path) if os.path.isfile(os.path.join(folder_path, filename))]
    times =[]
    results = {}
    
    for file_path in file_paths:
        kwargs = {
            #General
            'file_path': file_path, #MUST: Image or pdf path to OCR
            'binary_thres': 127, #Pixel value (0-255) to detect contourns, i.e, identify rectangles in the image
            'language': 'eng', #Language of the drawing, require installation of tesseract speficic language if not english
            'autoframe' : False, #Do we want to spot a frame as the maximum rectangle?
            'frame_thres': 0.9, #Frame boundary in % of img, if autoframe, this setting is overruled
            #GD&T
            'recognizer_gdt': recognizer_gdt, #MUST: A Tuple with (gdt alphabet, model path)
            'GDT_thres': 0.02, #Maximum porcentage of the image area to consider a cluster of rectangles a GD&T box
            #Dimensions
            'dimension_tuple': (detector, recognizer_dim, alphabet_dim), #MUST: A Tuple with (detector, dimension alphabet, model path)
            'cluster_thres': 20, #Minimum distance in pixels between two text predictions to consider the same text box
            'max_char': 15, #Max number of characters to consider a text prediction a dimension, otherwise -> other info
            'max_img_size': 1024, #Max size after applying scale for the img patch, bigger, better prediction, but more computationally expensive
            #Output
            'backg_save': False, #Option to save the background once all text and boxes have been removed, for synth training purposes
            'output_path': 'shit/', #Output path
            'save_mask': True, #Option to save the mask output
            'save_raw_output': True, #Option to save raw ouput, i.e, OCR text and box position,
            }
        
        results_, times_, _ = ocr_drawing(**kwargs)
        results[os.path.basename(file_path)]= results_
        times.append(sum(times_))
        print(
            "OCR in {}:\n"
            "Drawing segmentation:      {:.3f} s\n"
            "Table prediction:          {:.3f} s\n"
            "GD&T prediction:           {:.3f} s\n"
            "Dimension & other info:    {:.3f} s\n"
            "Saving & processing:       {:.3f} s\n"
            "----------------------\n"
            "Total time:                {:.3f} s\n"
            .format(os.path.basename(file_path),times_[0], times_[1], times_[2], times_[3], times_[4], sum(times_)))

    final_time = time.time()
    print(
    "Session Timing Report:\n"
    "Loading session:           {:.3f} s\n"
    "----------------------\n"
    .format(end_time - start_time))
    for i in range(len(file_paths)):
        print("OCR in {}: {:.3f} s".format(os.path.basename(file_paths[i]), times[i]))
    print(
        "----------------------\n"
        "Total time:           {:.3f} s\n"
        .format(final_time - start_time))
    return results