Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Getting Warning: skip broken file #1139

Closed
balandongiv opened this issue Jul 6, 2022 · 10 comments
Closed

Getting Warning: skip broken file #1139

balandongiv opened this issue Jul 6, 2022 · 10 comments

Comments

@balandongiv
Copy link
Contributor

balandongiv commented Jul 6, 2022 •
edited
Loading

While train the dbnetpp, the compiler return Warning: skip broken file . Unfortunately, this happen for almost all the images. Given limited dataset, I try my best to utilise max images possible.

May I know how to resolve this issue?

The full trace-back is as below:


after_val_epoch:
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
after_run:
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
2022-07-06 10:35:40,373 - mmocr - INFO - workflow: [('train', 1)], max: 6 epochs
2022-07-06 10:35:40,377 - mmocr - INFO - Checkpoints will be saved to /content/wdr/train_detect/base_dbnetpp by HardDiskBackend.
prepare index 13137 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N1_DG_BT.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1079, 558, 140, 95], 'segmentation': [[1079, 558, 1219, 558, 1219, 653, 1079, 653]], 'text': 'N1'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1073, 1005, 155, 95], 'segmentation': [[1073, 1005, 1228, 1005, 1228, 1100, 1073, 1100]], 'text': 'DG'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1070, 1120, 147, 95], 'segmentation': [[1070, 1120, 1217, 1120, 1217, 1215, 1070, 1215]], 'text': 'BT'}]} with img_prefix /content/drive/MyDrive/focr_dataset_all/batch_1
prepare index 1306 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N0_AG_EP.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [968, 678, 156, 96], 'segmentation': [[968, 678, 1124, 678, 1124, 774, 968, 774]], 'text': 'N0'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [962, 1125, 156, 97], 'segmentation': [[962, 1125, 1118, 1125, 1118, 1222, 962, 1222]], 'text': 'AG'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [962, 1240, 157, 95], 'segmentation': [[962, 1240, 1119, 1240, 1119, 1335, 962, 1335]], 'text': 'EP'}]} with img_prefix /content/drive/MyDrive/focr_dataset_all/batch_1
prepare index 7720 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N2_AG_Cd.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1060, 571, 157, 91], 'segmentation': [[1060, 571, 1217, 571, 1217, 662, 1060, 662]], 'text': 'N2'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1057, 1018, 152, 89], 'segmentation': [[1057, 1018, 1209, 1018, 1209, 1107, 1057, 1107]], 'text': 'AG'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1051, 1129, 152, 99], 'segmentation': [[1051, 1129, 1203, 1129, 1203, 1228, 1051, 1228]], 'text': 'Cd'}]} with img_prefix /content/drive/MyDrive/focr_dataset_all/batch_1
prepare index 14834 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N0_AG_GL.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1056, 683, 157, 94], 'segmentation': [[1056, 683, 1213, 683, 1213, 777, 1056, 777]], 'text': 'N0'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1052, 1131, 154, 96], 'segmentation': [[1052, 1131, 1206, 1131, 1206, 1227, 1052, 1227]], 'text': 'AG'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1051, 1245, 151, 97], 'segmentation': [[1051, 1245, 1202, 1245, 1202, 1342, 1051, 1342]], 'text': 'GL'}]} with img_prefix /content/drive/MyDrive/focr_dataset_all/batch_1
prepare index 11128 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'NI_DG_Bf.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1036, 507, 139, 92], 'segmentation': [[1036, 507, 1175, 507, 1175, 599, 1036, 599]], 'text': 'N1'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1031, 951, 144, 97], 'segmentation': [[1031, 951, 1175, 951, 1175, 1048, 1031, 1048]], 'text': 'DG'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1026, 1067, 139, 92], 'segmentation': [[1026, 1067, 1165, 1067, 1165, 1159, 1026, 1159]], 'text': 'Bf'}]} with img_prefix /content/drive/MyDrive/focr_dataset_all/batch_1
prepare index 4982 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N2_AB_JP.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1096, 675, 155, 97], 'segmentation': [[1096, 675, 1251, 675, 1251, 772, 1096, 772]], 'text': 'N2'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1090, 1123, 151, 94], 'segmentation': [[1090, 1123, 1241, 1123, 1241, 1217, 1090, 1217]], 'text': 'AB'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1085, 1240, 149, 96], 'segmentation': [[1085, 1240, 1234, 1240, 1234, 1336, 1085, 1336]], 'text': 'JP'}]} with img_prefix /content/drive/MyDrive/focr_dataset_all/batch_1
prepare index 2030 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N0_AG_EX.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1122, 572, 156, 95], 'segmentation': [[1122, 572, 1278, 572, 1278, 667, 1122, 667]], 'text': 'N0'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1120, 1019, 156, 93], 'segmentation': [[1120, 1019, 1276, 1019, 1276, 1112, 1120, 1112]], 'text': 'AG'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1121, 1136, 151, 93], 'segmentation': [[1121, 1136, 1272, 1136, 1272, 1229, 1121, 1229]], 'text': 'EX'}]} with img_prefix /content/drive/MyDrive/focr_dataset_all/batch_1
prepare index 6004 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N0_DH_P6.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1040, 544, 151, 93], 'segmentation': [[1040, 544, 1191, 544, 1191, 637, 1040, 637]], 'text': 'N0'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1042, 990, 149, 96], 'segmentation': [[1042, 990, 1191, 990, 1191, 1086, 1042, 1086]], 'text': 'DH'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1037, 1105, 155, 94], 'segmentation': [[1037, 1105, 1192, 1105, 1192, 1199, 1037, 1199]], 'text': 'P6'}]} with img_prefix /content/drive/MyDrive/focr_dataset_all/batch_1

In addition, store the image directly under colab directory produce similar issue

 -------------------- 
after_run:
(VERY_LOW    ) TextLoggerHook                     
 -------------------- 
2022-07-06 13:18:13,457 - mmocr - INFO - workflow: [('train', 1)], max: 6 epochs
2022-07-06 13:18:13,460 - mmocr - INFO - Checkpoints will be saved to /content/wdr/train_detect/base_dbnetpp by HardDiskBackend.

prepare index 100706 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N0,CB,Am_1.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1110, 812, 154, 97], 'segmentation': [[1110, 812, 1264, 812, 1264, 909, 1110, 909]], 'text': 'N0'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1110, 1260, 151, 97], 'segmentation': [[1110, 1260, 1261, 1260, 1261, 1357, 1110, 1357]], 'text': 'CB'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1108, 1375, 154, 97], 'segmentation': [[1108, 1375, 1262, 1375, 1262, 1472, 1108, 1472]], 'text': 'Am'}]} with img_prefix /content/batch2_v2
prepare index 207870 with error 'Polygon' object is not iterable
Warning: skip broken file {'file_name': 'N3,DA,Jm.png', 'height': 2048, 'width': 2448, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [1179, 752, 161, 104], 'segmentation': [[1179, 752, 1340, 752, 1340, 856, 1179, 856]], 'text': 'N3'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1188, 1199, 156, 102], 'segmentation': [[1188, 1199, 1344, 1199, 1344, 1301, 1188, 1301]], 'text': 'DA'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [1185, 1319, 157, 97], 'segmentation': [[1185, 1319, 1342, 1319, 1342, 1416, 1185, 1416]], 'text': 'Jm'}]} with img_prefix /content/batch2_v2
prepare index 155829 with error 'Polygon' object is not iterable
prepare index 33185 with error 'Polygon' object is not iterable

Is the issue related to here?
@Mountchicken
Copy link
Collaborator

Hi @balandongiv
It may be a problem with the dataset path. Please share us with you full configs,

@balandongiv
Copy link
Contributor Author

balandongiv commented Jul 7, 2022 •
edited
Loading

Thanks for replying @Mountchicken .

The images were stored as followed

#/content/batch2_v2
———batch2_v2
     |__ mmocr_compatible_annotation
     |__ img1.png
     |__ img1.json
     |__ img2.png
     |__ img2.json
     |__ img3.png
     |__ img3.json

and

#/content/batch2_v2/mmocr_compatible_annotation

———mmocr_compatible_annotation
     |__ crops ## contain crops images for txt recog
     |__ instances_training.txt
     |__ train_label.jsonl

The config is as below

_base_ = [
    '/content/mmocr/configs/_base_/default_runtime.py',
    '/content/mmocr/configs/_base_/schedules/schedule_sgd_1200e.py',
    '/content/mmocr/configs/_base_/det_models/dbnetpp_r50dcnv2_fpnc.py',
    '/content/mmocr/configs/_base_/det_pipelines/dbnet_pipeline.py',
]

# The YML suggested the DBNetpp is with Training Resources: 1x Nvidia A100
# Location where the annotation and crop images are being stored
root='/content/wdr'


# Set up working dir to save files and logs.
work_dir =f'{root}/train_detect/base_dbnetpp'


train_root_custm1 ='/content/batch2_v2'

train_custm1 = dict(            # This is the new one by
    type='TextDetDataset',
    img_prefix=train_root_custm1,
    ann_file=f'{train_root_custm1}/mmocr_compatible_annotation/instances_training.txt',
    loader=dict(
        type='AnnFileLoader',
        repeat=300,
        file_format='txt',
        parser=dict(
            type='LineJsonParser',
            keys=['file_name', 'height', 'width', 'annotations'])),
    pipeline=None,
    test_mode=False)


val_custm1 = dict(            # This is the new one by
    type='TextDetDataset',
    img_prefix=train_root_custm1,
    ann_file=f'{train_root_custm1}/mmocr_compatible_annotation/instances_training.txt',
    loader=dict(
        type='AnnFileLoader',
        repeat=1,
        file_format='txt',
        parser=dict(
            type='LineJsonParser',
            keys=['file_name', 'height', 'width', 'annotations'])),
    pipeline=None,
    test_mode=False)




train_list = [train_custm1]
test_list = [val_custm1]



train_pipeline_r50dcnv2 = {{_base_.train_pipeline_r50dcnv2}}
test_pipeline_4068_1024 = {{_base_.test_pipeline_4068_1024}}


data = dict(
    samples_per_gpu=16, # Default 32
    workers_per_gpu=8,
    val_dataloader=dict(samples_per_gpu=1),
    test_dataloader=dict(samples_per_gpu=1),
    train=dict(
        type='UniformConcatDataset',
        datasets=train_list,
        pipeline=train_pipeline_r50dcnv2),
    val=dict(
        type='UniformConcatDataset',
        datasets=test_list,
        pipeline=test_pipeline_4068_1024),
    test=dict(
        type='UniformConcatDataset',
        datasets=test_list,
        pipeline=test_pipeline_4068_1024))



evaluation = dict(
    interval=20,
    metric='hmean-iou')

Update

Remove the crop folder

#/content/batch2_v2/mmocr_compatible_annotation

———mmocr_compatible_annotation
     |__ instances_training.txt
     |__ train_label.jsonl

also result into similar Warning being thrown out

@balandongiv
Copy link
Contributor Author

Hi @Mountchicken , I notice this issue also happen when I am using the toy-dataset which can be reproduced via the following Notebook

In summary here the steps
First convert the coco to mmcor-compatible format via !python /mmocr/tools/data/common/labelme_converter.py /mmocr/tests/data/toy_dataset/labelme /mmocr/tests/data/toy_dataset/imgs/test /content --tasks recog --format jsonl

and with the following config

_base_ = [
    '/mmocr/configs/_base_/default_runtime.py',
    '/mmocr/configs/_base_/schedules/schedule_sgd_1200e.py',
    '/mmocr/configs/_base_/det_models/dbnetpp_r50dcnv2_fpnc.py',
    '/mmocr/configs/_base_/det_pipelines/dbnet_pipeline.py',
]
# /mmocr/configs/_base_/default_runtime.py
# The YML suggested the DBNetpp is with Training Resources: 1x Nvidia A100
# Location where the annotation and crop images are being stored
root='/content/wdr'


# Set up working dir to save files and logs.
work_dir =f'{root}/train_detect/base_dbnetpp'


train_root_custm1 ='/mmocr/tests/data/toy_dataset/imgs/test'

train_custm1 = dict(            # This is the new one by
    type='TextDetDataset',
    img_prefix=train_root_custm1,
    ann_file='/content/instances_training.txt',
    loader=dict(
        type='AnnFileLoader',
        repeat=300,
        file_format='txt',
        parser=dict(
            type='LineJsonParser',
            keys=['file_name', 'height', 'width', 'annotations'])),
    pipeline=None,
    test_mode=False)


val_custm1 = dict(            # This is the new one by
    type='TextDetDataset',
    img_prefix=train_root_custm1,
    ann_file=f'/content/instances_training.txt',
    loader=dict(
        type='AnnFileLoader',
        repeat=1,
        file_format='txt',
        parser=dict(
            type='LineJsonParser',
            keys=['file_name', 'height', 'width', 'annotations'])),
    pipeline=None,
    test_mode=False)




train_list = [train_custm1]
test_list = [val_custm1]



train_pipeline_r50dcnv2 = {{_base_.train_pipeline_r50dcnv2}}
test_pipeline_4068_1024 = {{_base_.test_pipeline_4068_1024}}


data = dict(
    samples_per_gpu=16, # Default 32
    workers_per_gpu=8,
    val_dataloader=dict(samples_per_gpu=1),
    test_dataloader=dict(samples_per_gpu=1),
    train=dict(
        type='UniformConcatDataset',
        datasets=train_list,
        pipeline=train_pipeline_r50dcnv2),
    val=dict(
        type='UniformConcatDataset',
        datasets=test_list,
        pipeline=test_pipeline_4068_1024),
    test=dict(
        type='UniformConcatDataset',
        datasets=test_list,
        pipeline=test_pipeline_4068_1024))



evaluation = dict(
    interval=20,
    metric='hmean-iou')

The train the model produce the following warning/error

2022-07-08 01:26:25,505 - mmocr - INFO - workflow: [('train', 1)], max: 6 epochs 2022-07-08 01:26:25,507 - mmocr - INFO - Checkpoints will be saved to /content/wdr/train_detect/base_dbnetpp by HardDiskBackend. prepare index 716 with error 'Polygon' object is not iterable prepare index 1040 with error 'Polygon' object is not iterable Warning: skip broken file {'file_name': 'img_1.jpg', 'height': 720, 'width': 1280, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [377, 117, 88, 13], 'segmentation': [[377, 117, 463, 117, 465, 130, 378, 130]], 'text': 'Genaxis Theatre'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [493, 115, 26, 16], 'segmentation': [[493, 115, 519, 115, 519, 131, 493, 131]], 'text': '[06]'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [374, 155, 35, 15], 'segmentation': [[374, 155, 409, 155, 409, 170, 374, 170]], 'text': '###'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [492, 151, 59, 19], 'segmentation': [[492, 151, 551, 151, 551, 170, 492, 170]], 'text': '62-03'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [376, 198, 46, 14], 'segmentation': [[376, 198, 422, 198, 422, 212, 376, 212]], 'text': 'Carpark'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [494, 189, 45, 17], 'segmentation': [[494, 190, 539, 189, 539, 205, 494, 206]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [372, 0, 122, 86], 'segmentation': [[374, 1, 494, 0, 492, 85, 372, 86]], 'text': '###'}]} with img_prefix /mmocr/tests/data/toy_dataset/imgs/test Warning: skip broken file {'file_name': 'img_3.jpg', 'height': 720, 'width': 1280, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [58, 71, 136, 52], 'segmentation': [[58, 80, 191, 71, 194, 114, 61, 123]], 'text': 'fusionopolis'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [147, 21, 29, 15], 'segmentation': [[147, 21, 176, 21, 176, 36, 147, 36]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [326, 75, 65, 38], 'segmentation': [[328, 75, 391, 81, 387, 112, 326, 113]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [401, 76, 47, 35], 'segmentation': [[401, 76, 448, 84, 445, 108, 402, 111]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [780, 6, 236, 36], 'segmentation': [[780, 7, 1015, 6, 1016, 37, 788, 42]], 'text': '###'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [221, 72, 91, 46], 'segmentation': [[221, 72, 311, 80, 312, 117, 222, 118]], 'text': 'fusionopolis'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [113, 19, 31, 14], 'segmentation': [[113, 19, 144, 19, 144, 33, 113, 33]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [257, 28, 51, 29], 'segmentation': [[257, 28, 308, 28, 308, 57, 257, 57]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [140, 115, 56, 18], 'segmentation': [[140, 120, 196, 115, 195, 129, 141, 133]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [86, 176, 26, 20], 'segmentation': [[86, 176, 110, 177, 112, 189, 89, 196]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [101, 185, 31, 19], 'segmentation': [[101, 193, 129, 185, 132, 198, 103, 204]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [223, 150, 71, 47], 'segmentation': [[223, 175, 244, 150, 294, 183, 235, 197]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [140, 232, 36, 24], 'segmentation': [[140, 239, 174, 232, 176, 247, 142, 256]], 'text': '###'}]} with img_prefix /mmocr/tests/data/toy_dataset/imgs/test prepare index 2214 with error 'Polygon' object is not iterable prepare index 1109 with error 'Polygon' object is not iterable prepare index 2149 with error 'Polygon' object is not iterable Warning: skip broken file {'file_name': 'img_5.jpg', 'height': 720, 'width': 1280, 'annotations': [{'iscrowd': 1, 'category_id': 1, 'bbox': [405, 409, 32, 52], 'segmentation': [[408, 409, 437, 436, 434, 461, 405, 433]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [435, 434, 8, 33], 'segmentation': [[437, 434, 443, 440, 441, 467, 435, 462]], 'text': '###'}]} with img_prefix /mmocr/tests/data/toy_dataset/imgs/test Warning: skip broken file {'file_name': 'img_2.jpg', 'height': 720, 'width': 1280, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [602, 173, 33, 24], 'segmentation': [[602, 173, 635, 175, 634, 197, 602, 196]], 'text': 'EXIT'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [734, 310, 58, 54], 'segmentation': [[734, 310, 792, 320, 792, 364, 738, 361]], 'text': 'I2R'}]} with img_prefix /mmocr/tests/data/toy_dataset/imgs/test Warning: skip broken file {'file_name': 'img_2.jpg', 'height': 720, 'width': 1280, 'annotations': [{'iscrowd': 0, 'category_id': 1, 'bbox': [602, 173, 33, 24], 'segmentation': [[602, 173, 635, 175, 634, 197, 602, 196]], 'text': 'EXIT'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [734, 310, 58, 54], 'segmentation': [[734, 310, 792, 320, 792, 364, 738, 361]], 'text': 'I2R'}]} with img_prefix /mmocr/tests/data/toy_dataset/imgs/test prepare index 1171 with error 'Polygon' object is not iterable Warning: skip broken file {'file_name': 'img_10.jpg', 'height': 720, 'width': 1280, 'annotations': [{'iscrowd': 1, 'category_id': 1, 'bbox': [260, 138, 24, 20], 'segmentation': [[261, 138, 284, 140, 279, 158, 260, 158]], 'text': '###'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [288, 138, 129, 23], 'segmentation': [[288, 138, 417, 140, 416, 161, 290, 157]], 'text': 'HarbourFront'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [743, 145, 37, 18], 'segmentation': [[743, 145, 779, 146, 780, 163, 746, 163]], 'text': 'CC22'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [783, 129, 50, 26], 'segmentation': [[783, 129, 831, 132, 833, 155, 785, 153]], 'text': 'bua'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [831, 133, 43, 23], 'segmentation': [[831, 133, 870, 135, 874, 156, 835, 155]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [159, 204, 72, 15], 'segmentation': [[159, 205, 230, 204, 231, 218, 159, 219]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [785, 158, 75, 21], 'segmentation': [[785, 158, 856, 158, 860, 178, 787, 179]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [1011, 157, 68, 16], 'segmentation': [[1011, 157, 1079, 160, 1076, 173, 1011, 170]], 'text': '###'}]} with img_prefix /mmocr/tests/data/toy_dataset/imgs/test prepare index 948 with error 'Polygon' object is not iterable Warning: skip broken file {'file_name': 'img_7.jpg', 'height': 720, 'width': 1280, 'annotations': [{'iscrowd': 1, 'category_id': 1, 'bbox': [345, 130, 56, 23], 'segmentation': [[346, 133, 400, 130, 401, 148, 345, 153]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [301, 123, 50, 35], 'segmentation': [[301, 127, 349, 123, 351, 154, 303, 158]], 'text': '###'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [869, 61, 54, 30], 'segmentation': [[869, 67, 920, 61, 923, 85, 872, 91]], 'text': 'citi'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [884, 141, 50, 19], 'segmentation': [[886, 144, 934, 141, 932, 157, 884, 160]], 'text': 'smrt'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [634, 86, 182, 35], 'segmentation': [[634, 106, 812, 86, 816, 104, 634, 121]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [418, 112, 53, 36], 'segmentation': [[418, 117, 469, 112, 471, 143, 420, 148]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [634, 107, 149, 28], 'segmentation': [[634, 124, 781, 107, 783, 123, 635, 135]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [634, 117, 210, 38], 'segmentation': [[634, 138, 844, 117, 843, 141, 636, 155]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [468, 117, 57, 26], 'segmentation': [[468, 124, 518, 117, 525, 138, 468, 143]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [301, 162, 231, 39], 'segmentation': [[301, 181, 532, 162, 530, 182, 301, 201]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [296, 147, 104, 27], 'segmentation': [[296, 157, 396, 147, 400, 165, 300, 174]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [420, 136, 107, 27], 'segmentation': [[420, 151, 526, 136, 527, 154, 421, 163]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [616, 250, 41, 35], 'segmentation': [[617, 251, 657, 250, 656, 282, 616, 285]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [695, 243, 43, 35], 'segmentation': [[695, 246, 738, 243, 738, 276, 698, 278]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [739, 241, 24, 21], 'segmentation': [[739, 241, 760, 241, 763, 260, 742, 262]], 'text': '###'}]} with img_prefix /mmocr/tests/data/toy_dataset/imgs/test prepare index 1017 with error 'Polygon' object is not iterable Warning: skip broken file {'file_name': 'img_6.jpg', 'height': 720, 'width': 1280, 'annotations': [{'iscrowd': 1, 'category_id': 1, 'bbox': [875, 92, 35, 20], 'segmentation': [[875, 92, 910, 92, 910, 112, 875, 112]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [748, 95, 39, 14], 'segmentation': [[748, 95, 787, 95, 787, 109, 748, 109]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [106, 394, 47, 31], 'segmentation': [[106, 395, 150, 394, 153, 425, 106, 424]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [165, 393, 48, 28], 'segmentation': [[165, 393, 213, 396, 210, 421, 165, 421]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [705, 49, 42, 15], 'segmentation': [[706, 52, 747, 49, 746, 62, 705, 64]], 'text': '###'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [111, 459, 96, 23], 'segmentation': [[111, 459, 206, 461, 207, 482, 113, 480]], 'text': 'Reserve'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [831, 9, 63, 13], 'segmentation': [[831, 9, 894, 9, 894, 22, 831, 22]], 'text': '###'}, {'iscrowd': 0, 'category_id': 1, 'bbox': [641, 454, 52, 15], 'segmentation': [[641, 456, 693, 454, 693, 467, 641, 469]], 'text': 'CAUTION'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [839, 32, 52, 15], 'segmentation': [[839, 32, 891, 32, 891, 47, 839, 47]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [788, 46, 43, 13], 'segmentation': [[788, 46, 831, 46, 831, 59, 788, 59]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [830, 95, 42, 11], 'segmentation': [[830, 95, 872, 95, 872, 106, 830, 106]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [921, 92, 31, 19], 'segmentation': [[921, 92, 952, 92, 952, 111, 921, 111]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [968, 40, 45, 13], 'segmentation': [[968, 40, 1013, 40, 1013, 53, 968, 53]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [1002, 89, 29, 11], 'segmentation': [[1002, 89, 1031, 89, 1031, 100, 1002, 100]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [1043, 38, 55, 14], 'segmentation': [[1043, 38, 1098, 38, 1098, 52, 1043, 52]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [1069, 85, 69, 14], 'segmentation': [[1069, 85, 1138, 85, 1138, 99, 1069, 99]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [1128, 36, 50, 16], 'segmentation': [[1128, 36, 1178, 36, 1178, 52, 1128, 52]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [1168, 84, 32, 13], 'segmentation': [[1168, 84, 1200, 84, 1200, 97, 1168, 97]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [1219, 27, 40, 22], 'segmentation': [[1223, 27, 1259, 27, 1255, 49, 1219, 49]], 'text': '###'}, {'iscrowd': 1, 'category_id': 1, 'bbox': [1264, 28, 15, 18], 'segmentation': [[1264, 28, 1279, 28, 1279, 46, 1264, 46]], 'text': '###'}]} with img_prefix /mmocr/tests/data/toy_dataset/imgs/test

@Mountchicken
Copy link
Collaborator

@balandongiv
Sry for the late reply. I meet the same problem before on colab and it also throws the same error. This may still be a problem with the path. You could try putting the annotation files and images in one folder, like mmocr does。

|__root_dir
    |__imgs
         |__img1.jpg
         |__img2.jpg
    |__ instance_training.json

@balandongiv
Copy link
Contributor Author

Thanks for replying @Mountchicken , appreciate it (its really like waiting for Santa). anyhow, I did as recommend. but, the issues still persist and can be reproduce via this notebook.

The file/folder structure is as below

#/mmocr/tests/data/toy_dataset/imgs

———imgs
     |__ crops
     |__ test
         |__img1.jpg
         |__img2.jpg
     |__ instances_training.txt

The instance_training.txt content is as below

{"file_name": "img_3.jpg", "height": 720, "width": 1280, "annotations": [{"iscrowd": 0, "category_id": 1, "bbox": [58, 71, 136, 52], "segmentation": [[58, 80, 191, 71, 194, 114, 61, 123]], "text": "fusionopolis"}, {"iscrowd": 1, "category_id": 1, "bbox": [147, 21, 29, 15], "segmentation": [[147, 21, 176, 21, 176, 36, 147, 36]], "text": "###"}, {"iscrowd": 1, "category_id": 1, "bbox": [326, 75, 65, 38], "segmentation": [[328, 75, 391, 81, 387, 112, 326, 113]], "text": "###"}, {"iscrowd": 1, "category_id": 1, "bbox": [401, 76, 47, 35], "segmentation": [[401, 76, 448, 84, 445, 108, 402, 111]], "text": "###"}, {"iscrowd": 1, "category_id": 1, "bbox": [780, 6, 236, 36], "segmentation": [[780, 7, 1015, 6, 1016, 37, 788, 42]], "text": "###"}, {"iscrowd": 0, "category_id": 1, "bbox": [221, 72, 91, 46], "segmentation": [[221, 72, 311, 80, 312, 117, 222, 118]], "text": "fusionopolis"}, {"iscrowd": 1, "category_id": 1, "bbox": [113, 19,}

@gaotongxiao
Copy link
Collaborator

Have you tested it on a GPU machine? I am able to train the model with your config without any trouble on my machine, but on colab the process does fail on the transform. Maybe there is some minor difference we were not able to realize.

@balandongiv
Copy link
Contributor Author

Thanks for responding @gaotongxiao., Right now, I am having issue installing the mmcv-full in my local machine hence the desperation to train the model under Colab.

May I know what do you mean by

some minor difference we were not able to realize

Is it from my side or the setting within the mmocr?

@gaotongxiao
Copy link
Collaborator

I'd suggest you install mmcv-full via mim. First make sure you have PyTorch installed, then:

pip install mim
mim install mmcv-full

And mim should find the right mmcv-full package that works best for your configuration.

May I know what do you mean by

some minor difference we were not able to realize

Is it from my side or the setting within the mmocr?

I was not able to identify the reason as well, and the only clue that I have is that your config actually works well locally.

@balandongiv
Copy link
Contributor Author

balandongiv commented Jul 12, 2022 •
edited
Loading

Thanks for replying @gaotongxiao ,

Actually I have installed using the recommended approach. But I got an Getting RuntimeError: box_iou_rotated_impl : implementation for device cuda:0 not found when running the test file, as I reported here .

Similarly, when running the the following in my local

_base_ = [
    'configs/_base_/default_runtime.py',
    'configs/_base_/schedules/schedule_sgd_1200e.py',
    'configs/_base_/det_models/dbnetpp_r50dcnv2_fpnc.py',
    'configs/_base_/det_pipelines/dbnet_pipeline.py',
]

it produce

RuntimeError: modulated_deformable_im2col_impl: implementation for device cuda:0 not found.

I think the issue boil up at implementation for device cuda:0 not found

open-mmlab/mmcv#2110

@balandongiv
Copy link
Contributor Author

Hi @gaotongxiao & @Mountchicken ,

I now can confirm the only occur in Google Colab but not in local machine. Thanks for your time

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@balandongiv @gaotongxiao @Mountchicken