OCR0056: Extracting Line Bounding Boxes Details and Line Segments from Synthetic Page Images

Description:

This project focuses on enhancing the existing package that generates synthetic page images for Tibetan Pecha and Modern Tibetan Book formats. The main tasks involve modifying the scripts (pecha_format_page.py and modern_book_format_page.py) to extract polygonal bounding box information tightly around each line of text on the synthetic pages. These tight bounding boxes are crucial for creating training data for line segmentation models. Additionally, the project will apply an offset to these bounding boxes to extract individual lines more realistically for OCR model training. The extracted lines, along with their transcripts, will be organized and saved in a structured format to develop and improve OCR models effectively.

Completion Criteria:

Ensure the modified scripts (pecha_format_page.py and modern_book_format_page.py) accurately extract polygonal bounding box information for each line on synthetic page images.
Save the extracted polygonal bounding box details for each page in a specified JSONL format, ensuring they are tight around the text lines.
Apply an offset to the polygonal bounding boxes to make the extracted lines more realistic, and extract individual lines from both page formats (Tibetan Pecha and Modern Tibetan Book), including augmented and non-augmented versions.
Generate and save corresponding transcripts for each extracted line.
Organize all extracted lines and transcripts systematically for training OCR models.

Method Applicable For Augmentation List :

1) BadPhotoCopy 2) Brightness 3) Contrast 4) Blur 5) Background 6) Grid Distort 7) InkBleed 8) RandomShadow

Current Method Not Applicable for Augmentation List:

1) Rotation 2) Deform 3) Scribble 4) SunFlare 5) DirtySpot 6) Low Ink

Implementation:

Subtasks:

[x] Modify Synthetic Page Generation Scripts:
- [x] Update pecha_format_page.py and modern_book_format_page.py to generate synthetic pages for both Tibetan Pecha and Modern Tibetan Book formats.
- [x] Add functionality to extract polygonal bounding box details for each line on the synthetic pages while generating them.
- [x] Visualize the extracted polygonal bounding boxes to ensure they are tight around the text lines (for line segmentation model training).
- [x] Save the validated polygonal bounding box details for each page in JSONL format after confirming they are tight.
[ ] Modify Line Extraction Script (extraction_line.py):
- [ ] Update the script to read polygonal bounding box information from the JSONL files.
- [ ] Apply an offset to the polygonal bounding boxes before extracting the lines to make them more realistic for OCR model training.
- [ ] Extract lines based on the adjusted bounding boxes and save them along with their corresponding transcripts for both formats.
- [ ] Organize and save the extracted data properly for use in training OCR models.

Initial Estimation start date: 29/08/24 end date: 3/09/24

Update Estimation Start date: 4/09/24 End date: 5/09/24

@devatwiai and @ta4tsering, could you please review my card? It involves extracting line bounding boxes details for each line on synthetic pages, saving this data, and using it to extract lines from each synthetic page while generating corresponding transcripts for each line.

Visualization of the Tight Bounding Box For each Line on Page Image

page_1_1123x265_count_1_font30_Kangba Derchi-Drutsa

Issues:

page_1_1800x630_count_1_font30_Tendar Dutsa

Applying Morphological Operations with Contour Detection

page_1_1800x630_count_1_font30_Kangba Derchi-Drutsa

Hi @devatwiai, After evaluating various methods, I’ve found that Morphological Operations with Contour Detection is the most effective for extracting bounding boxes from clean, synthetic images. This approach is both fast and efficient.

Highlights: 1) Efficiency: This method is ideal for noise-free synthetic images, providing quick and accurate results. 2) Tight Bounding Boxes: Using approxPolyDP with an adjustable epsilon parameter allows for precise control over bounding box tightness. 3) Flexibility: It adapts well to different text sizes and configurations, ensuring robust detection.

However, I’ve also come across models like EAST (Efficient and Accurate Scene Text detector). While I’m not certain if EAST will provide tighter bounding boxes compared to the morphological operations method, it’s worth considering if the text detection conditions become more complex. Given that we are currently working with noise-free synthetic images, Morphological Operations seems to be the best fit. Nonetheless, exploring the EAST model might be beneficial if you anticipate dealing with more varied or challenging text detection scenarios.

Line Bounding Box Details Example

{"id": "page_76_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_76_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "height": 57, "width": 1022, "center": [549.0, 198.5], "points": [[38, 170], [38, 227], [1060, 227], [1060, 170]]}
{"id": "page_76_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_76_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "height": 57, "width": 1052, "center": [556.0, 139.5], "points": [[30, 111], [30, 168], [1082, 168], [1082, 111]]}
{"id": "page_76_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_76_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "height": 57, "width": 1039, "center": [558.5, 81.5], "points": [[39, 53], [39, 110], [1078, 110], [1078, 53]]}

Working TIghter Bounding Box For Augmented Page

page_2_1123x265_count_1_font30_玉翅+zhuca+weizang_DeformAugmentation page_3_1800x630_count_1_font30_玉翅+zhuca+weizang_RotationAugmentation(angle=9 94)_DeformAugmentation

Issue on Blending the text into the background

page_68_1680x402_count_9_font30_玉翅+zhuca+weizang_ScribbleAugmentation

@devatwiai the tight bounding box working fine with the noise page image

Points Details in Coordinates of Polygon Bounding box

{"id": "page_25_1000x128_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_25_1000x128_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "height": 45.0, "width": 125.0, "center": [451.5, 85.5], "points": [[394, 66], [396, 100], [389, 103], [395, 103], [396, 108], [414, 106], [426, 90], [420, 89], [422, 83], [431, 82], [441, 105], [449, 105], [446, 74], [454, 103], [462, 103], [462, 93], [475, 100], [478, 106], [486, 106], [504, 96], [505, 108], [513, 108], [514, 64], [450, 64], [445, 69], [440, 64], [412, 63], [408, 68]]}
{"id": "page_25_1000x128_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_25_1000x128_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "height": 49.0, "width": 36.0, "center": [334.0, 83.5], "points": [[351, 64], [344, 64], [343, 59], [335, 59], [331, 63], [317, 64], [316, 108], [324, 108], [327, 84], [332, 92], [335, 108], [350, 108], [352, 96]]}
{"id": "page_25_1000x128_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_25_1000x128_count_1_font30_\u7389\u7fc5+zhuca+weizang.png", "height": 26.455835342407227, "width": 49.47584915161133, "center": [64.3785400390625, 83.61317443847656], "points": [[60, 59], [56, 63], [51, 79], [57, 92], [60, 108], [76, 108], [78, 96], [77, 64], [69, 64]]}

Blending the text with the background:

page_12_1680x402_count_2_font30_玉翅+zhuca+weizang_ContrastAugmentation(factor= 1 25)_ScribbleAugmentation

Update on Bounding Box Details:


{"id": "page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_0", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_0", "height": 57, "width": 993, "center": [548.5, 198.5], "points": [[52, 170], [52, 227], [1045, 227], [1045, 170]], "page_number": 1}
{"id": "page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_1", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_1", "height": 57, "width": 1023, "center": [555.5, 139.5], "points": [[44, 111], [44, 168], [1067, 168], [1067, 111]], "page_number": 1}
{"id": "page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_2", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_2", "height": 51, "width": 23, "center": [64.5, 84.5], "points": [[53, 59], [53, 110], [76, 110], [76, 59]], "page_number": 1}
{"id": "page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_3", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_3", "height": 53, "width": 238, "center": [231.0, 83.5], "points": [[112, 57], [112, 110], [350, 110], [350, 57]], "page_number": 1}
{"id": "page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_4", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/page_1_1123x265_count_1_font30_\u7389\u7fc5+zhuca+weizang.png_bbox_4", "height": 56, "width": 669, "center": [728.5, 81.0], "points": [[394, 53], [394, 109], [1063, 109], [1063, 53]], "page_number": 1}```

Extracting Green Bounding Box Pixel Details from the augmented page image with bounding box

original_image_with_contours

visualized_bounding_box_contours

In Jsonl format for the bounding box details of the Augmented Page


{"id": "with_bounding_box_aug_page_24_1123x265_count_9_font30_\u7389\u7fc5+zhuca+weizang_aug_Def.png_polygon_bbox_0", "image": "https://s3.amazonaws.com/monlam.ai.ocr/line_segmentations/Images/with_bounding_box_aug_page_24_1123x265_count_9_font30_\u7389\u7fc5+zhuca+weizang_aug_Def.png_polygon_bbox_0", "points": [[1035, 167], [1034, 168], [1032, 168], [1031, 169], [1030, 169], [1028, 171], [1021, 171], [1020, 172], [1007, 172], [1006, 173], [993, 173], [992, 174], [983, 174], [982, 175], [976, 175], [973, 178], [979, 178], [982, 181], [981, 182], [976, 182], [975, 183], [969, 183], [968, 184], [962, 184], [961, 185], [955, 185], [954, 186], [948, 186], [947, 187], [941, 187], [940, 188], [939, 187], [936, 187], [935, 186], [932, 186], [931, 185], [929, 185], [927, 183], [927, 182], [926, 181], [926, 179], [925, 178], [926, 179], [926, 181], [927, 182], [927, 183], [928, 184], [928, 185], [931, 185], [932, 186], [934, 186], [935, 187], [938, 187], [939, 188], [941, 188], [942, 187], [948, 187], [949, 186], [956, 186], [957, 185], [964, 185], [965, 184], [970, 184], [971, 183], [977, 183], [978, 182], [981, 182], [982, 181], [979, 178], [978, 178], [977, 177], [977, 176], [978, 175], [984, 175], [985, 174], [995, 174], [996, 173], [997, 174], [998, 173], [1009, 173], [1010, 172], [1023, 172], [1024, 171], [1028, 171], [1030, 169], [1031, 169], [1032, 168], [1034, 168], [1035, 167], [1036, 167], [1037, 168], [1040, 168], [1041, 169], [1044, 169], [1045, 170], [1048, 170], [1049, 171], [1052, 171], [1053, 172], [1056, 172], [1057, 173], [1060, 173], [1061, 174], [1064, 174], [1065, 175], [1067, 175], [1069, 177], [1069, 179], [1072, 182], [1072, 184], [1073, 185], [1073, 188], [1074, 189], [1076, 189], [1077, 190], [1080, 190], [1081, 191], [1081, 197], [1082, 198], [1082, 199], [1083, 200], [1083, 206], [1082, 207], [1078, 207], [1077, 206], [1076, 207], [1076, 216], [1075, 217], [1075, 222], [1074, 223], [1074, 226], [1073, 227], [1070, 227], [1069, 226], [1066, 226], [1065, 225], [1062, 225], [1061, 224], [1059, 224], [1058, 223], [1054, 223], [1053, 222], [1050, 222], [1049, 221], [1046, 221], [1045, 220], [1042, 220], [1041, 219], [1038, 219], [1037, 218], [1021, 218], [1020, 219], [1007, 219], [1006, 220], [993, 220], [992, 221], [983, 221], [982, 222], [976, 222], [975, 223], [969, 223], [968, 224], [967, 224], [965, 226], [962, 226], [961, 227], [955, 227], [954, 228], [948, 228], [947, 229], [937, 229], [935, 227], [932, 227], [931, 226], [929, 226], [928, 225], [925, 225], [924, 224], [922, 224], [921, 223], [918, 223], [917, 222], [915, 222], [914, 221], [911, 221], [910, 220], [908, 220], [907, 219], [904, 219], [903, 218], [901, 218], [900, 217], [897, 217], [896, 216], [889, 216], [888, 217], [885, 217], [885, 218], [884, 219], [882, 219], [881, 220], [878, 220], [877, 221], [874, 221], [873, 222], [871, 222], [870, 223], [867, 223], [866, 224], [863, 224], [862, 225], [860, 225], [859, 226], [855, 226], [853, 224], [852, 224], [851, 225], [849, 225], [848, 226], [840, 226], [839, 225], [832, 225], [831, 224], [824, 224], [823, 223], [815, 223], [814, 222], [807, 222], [806, 221], [799, 221], [798, 220], [790, 220], [789, 219], [778, 219], [777, 218], [768, 218], [767, 219], [766, 218], [765, 219], [756, 219], [755, 218], [750, 218], [749, 219], [746, 219], [745, 220], [743, 220], [742, 221], [739, 221], [738, 222], [736, 222], [735, 221], [734, 222], [733, 222], [732, 221], [731, 221], [730, 222], [729, 222], [727, 220], [727, 219], [728, 218], [729, 218], [734, 213], [735, 213], [735, 212], [739, 208], [739, 207], [740, 206], [740, 205], [741, 204], [741, 202], [742, 201], [742, 199], [741, 198], [739, 198], [738, 199], [735, 199], [734, 200], [731, 200], [730, 201], [727, 201], [726, 202], [724, 202], [723, 203], [720, 203], [719, 204], [716, 204], [715, 205], [712, 205], [711, 206], [708, 206], [707, 207], [702, 207], [701, 206], [699, 206], [698, 205], [695, 205], [694, 204], [692, 204], [691, 203], [688, 203], [687, 202], [684, 202], [684, 204], [685, 205], [685, 210], [686, 211], [686, 218], [687, 219], [687, 222], [687, 219], [686, 218], [686, 211], [685, 210], [685, 205], [684, 204], [684, 203], [685, 202], [687, 202], [688, 203], [690, 203], [691, 204], [693, 204], [694, 205], [697, 205], [698, 206], [700, 206], [701, 207], [708, 207], [709, 206], [712, 206], [713, 205], [716, 205], [717, 204], [720, 204], [721, 203], [723, 203], [724, 202], [727, 202], [728, 201], [731, 201], [732, 200], [735, 200], [736, 199], [739, 199], [740, 198], [741, 198], [742, 199], [742, 201], [741, 202], [741, 204], [740, 205], [740, 206], [739, 207], [739, 208], [732, 215], [731, 215], [731, 216], [729, 218], [728, 218], [726, 220], [725, 220], [723, 222], [722, 222], [721, 223], [720, 223], [719, 224], [720, 224], [721, 223], [723, 223], [724, 222], [727, 222], [728, 221], [729, 222], [731, 222], [732, 221], [733, 222], [739, 222], [740, 221], [743, 221], [744, 220], [746, 220], [747, 219], [750, 219], [751, 218], [753, 218], [754, 219], [764, 219], [765, 220], [766, 219], [767, 219], [768, 218], [774, 218], [775, 219], [785, 219], [786, 220], [795, 220], [796, 221], [804, 221], [805, 222], [812, 222], [813, 223], [821, 223], [822, 224], [829, 224], [830, 225], [837, 225], [838, 226], [848, 226], [849, 225], [852, 225], [853, 224], [855, 226], [855, 227], [856, 227], [857, 226], [859, 226], [860, 225], [863, 225], [864, 224], [867, 224], [868, 223], [870, 223], [871, 222], [874, 222], [875, 221], [878, 221], [879, 220], [881, 220], [882, 219], [884, 219], [886, 217], [889, 217], [890, 216], [895, 216], [896, 217], [899, 217], [900, 218], [902, 218], [903, 219], [906, 219], [907, 220], [909, 220], [910, 221], [913, 221], [914, 222], [916, 222], [917, 223], [920, 223], [921, 224], [924, 224], [925, 225], [927, 225], [928, 226], [931, 226], [932, 227], [934, 227], [935, 228], [936, 228], [937, 229], [938, 229], [939, 230], [941, 230], [942, 229], [948, 229], [949, 228], [955, 228], [956, 227], [963, 227], [964, 226], [966, 226], [966, 225], [967, 224], [970, 224], [971, 223], [977, 223], [978, 222], [984, 222], [985, 221], [995, 221], [996, 220], [1009, 220], [1010, 219], [1023, 219], [1024, 218], [1036, 218], [1037, 219], [1040, 219], [1041, 220], [1044, 220], [1045, 221], [1048, 221], [1049, 222], [1052, 222], [1053, 223], [1056, 223], [1057, 224], [1060, 224], [1061, 225], [1064, 225], [1065, 226], [1068, 226], [1069, 227], [1072, 227], [1073, 228], [1074, 228], [1074, 223], [1075, 222], [1075, 217], [1076, 216], [1076, 208], [1077, 207], [1083, 207], [1084, 206], [1084, 205], [1083, 204], [1083, 199], [1081, 197], [1081, 191], [1082, 190], [1078, 190], [1077, 189], [1074, 189], [1073, 188], [1073, 185], [1072, 184], [1072, 182], [1069, 179], [1069, 177], [1067, 175], [1066, 175], [1065, 174], [1062, 174], [1061, 173], [1058, 173], [1057, 172], [1054, 172], [1053, 171], [1050, 171], [1049, 170], [1046, 170], [1045, 169], [1042, 169], [1041, 168], [1038, 168], [1037, 167]], "page_number": 1}```

Extracting line from the augmented page image using the green pixel coordinates(Polygon Bounding Box)

1) plain Augmented page Image:

1123x265_pecha_format_page_image

Line Extraction:

1123x265_pecha_format_page_image_without_bbox_line_5

1123x265_pecha_format_page_image_without_bbox_line_4

1123x265_pecha_format_page_image_without_bbox_line_3

2) Rotated page Image:

original_image_with_contours

Extracted line:

794x265_aug_withb_box_line_0

Issue on Extracted line:

794x265_aug_withb_box_line_2

Extracted lines :

line_2 line_3

Transcription:

ཊཚེ།༡༦།།༠།༢༥།༥ལིཀ༠༠༥༧༠༥༠༠༦/༠༠།༧༠༥༠༡༩༥གི༡༧༠་ི་ཞ༠༠༠༡ ྣརིྲགྣགིྣཾ།


་༡ི༣།༩༡༠༠༠ཧི༢ཞི༣།།།༥༠།༥༠ཞ༥།༡༢༩།༥:༩།༥༩ཀ"ེེ༩༠༩༥༠༥།༩།།ག༠༥༦༧༠༥།༥ཀེ༠།༥༠༥།༩།༥།།ག༥༦ན༥།ཉེན༥༢༧༧༢༤ུ༠༥༥༥༧ེད།༨༠༥༡༣༠༥༩༥༥ཞ་༥༦ེ༢༠༠༧༥། ཤའིཡི9།༠༡༣༡༥༥ ```

Synthetic page image

1123x265_pecha_format_page_image_without_bbox

Line Transcript

༄༅༅།     །རྒྱ་གར་སྐད་དུ། བི་ན་ཡ་བསྟུ། བོད་སྐད་དུ། འདུལ་བ་གཞི། བམ་པོ་དང་པོ། དཀོན་མཆོག་གསུམ་ལ་ཕྱག་འཚལ་ལོ། །གང་གིས་འཆིང་རྣམས་ཡང་དག་རབ་བཅད་ཅིང་། །མུ་སྟེགས་ཚོགས་རྣམས་ཐམས་ཅད་རབ་བཅོམ་སྟེ། །སྡེ་དང་བཅས་པའི་བདུད་རྣམས་ངེས་བཅོམ་ནས། །བྱང་ཆུབ་འདི་བརྙེས་དེ་ལ་ཕྱག་འཚལ་ལོ། །ཁྱིམ་

དོན་ཆེ་ཆུང་སྤངས་ཏེ་དང་པོར་རབ་འབྱུང་དཀའ། །རབ་བྱུང་ཐོབ་ནས་ཡུལ་སྤྱད་དག་གིས་དགའ་ཐོབ་དཀའ། །མངོན་དགའ་ཇི་བཞིན་དོན་བསྐྱེད་ཡང་དག་བྱེད་པ་དཀའ། །ངུར་སྨྲིག་གོས་འཆང་མཁས་པ་ཚུལ་ལས་ཉམས་པ་དཀའ། །གཞི་རྣམས་ཀྱི་སྤྱི་སྡོམ་ལ། རབ་འབྱུང་གསོ་སྦྱོང་གཞི་དང་ནི། །དགག་དབྱེ་དབྱར་དང་ཀོ་ལྤགས་གཞི། །སྨན་དང་གོས་དང་སྲ་བརྐྱང་དང་

། །ཀཽ་ཤཱམ་བཱི་དང་ལས་ཀྱི་གཞི། །དམར་སེར་ཅན་དང་གང་ཟག་དང་། །སྤོ་དང་གསོ་སྦྱོང་བཞག་པ་དང་། །གནས་མལ་དང་ནི་རྩོད་པ་དང་། །དགེ་འདུན་དབྱེན་རྣམས་བསྡུས་པ་ཡིན། །རབ་ཏུ་འབྱུང་བའི་གཞིའི་སྤྱི་སྡོམ་ལ། ཤཱ་རིའི་བུ་དང་མུ་སྟེགས་ཅན། །དགེ་ཚུལ་གཉིས་དང་བྱ་རོག་སྐྲོད། །དགྲ་བཅོམ་བསད་དང་ལག་རྡུམ་གྱི། །སྡེ་ཚན་ཡང་དག་བསྡུས་པ་ཡིན། །

སྡོམ་ལ། ཤཱ་རིའི་བུ་དང་རབ་འབྱུང་དང་། །བསྙེན་པར་རྫོགས་པར་གནང་བ་དང་། །ཉེ་སྡེས་ཚོགས་ནི་བསྡུས་པ་དང་། །ལྔ་པའི་སྡེ་ཚན་བསྡུས་པ་ཡིན། །བྱང་ཆུབ་སེམས་དཔའ་དགའ་ལྡན་གྱི་གནས་ན་བཞུགས་པ་ན།ཡུལ་ཨང་ག་དག་ན་ཨང་གའི་རྒྱལ་པོ་ཞེས་བྱ་བས་རྒྱལ་སྲིད་འབྱོར་པ། རྒྱས་པ་བདེ་བ་ལོ་ལེགས་པ་སྐྱེ་བོ་དང་མི་མང་པོས་གང་བ་

བྱེད་དུ་བཅུག་གོ། །ཡུལ་མ་ག་དྷཱ་དག་ན་ཡང་རྒྱལ་པོ་པད་མ་ཆེན་པོ་ཞེས་བྱ་བས། རྒྱལ་སྲིད་འབྱོར་པ་རྒྱས་པ་བདེ་བ་ལོ་ལེགས་པ་སྐྱེ་བོ་དང་མི་མང་པོས་གང་བ་བྱེད་དུ་བཅུག་གོ། །རེས་འགའ་ནི་ཨང་གའི་རྒྱལ་པོ་དཔུང་དང་མཐུ་ཆེ་བ་ཡིན་ལ། རེས་འགའ་ནི་རྒྱལ་པོ་པད་མ་ཆེན་པོ་དཔུང་དང་མཐུ་ཆེ་བ་ཡིན་ནོ། །གང་གི་ཚེ་ཨང་གའི་

OpenPecha / SynthImage