Link mentioned on page to download SynthText from baidu cloud gives 404 error. Either the link is expired or it has been moved. Can someone kindly fix that error.

Sharing with this issue on Pytorch 0.4.0

Traceback (most recent call last):
File "train_textsnake.py", line 241, in
main()
File "train_textsnake.py", line 226, in main
train(model, train_loader, criterion, scheduler, optimizer, epoch, logger)
File "train_textsnake.py", line 66, in train
for i, (img, train_mask, tr_mask, tcl_mask, radius_map, sin_map, cos_map, meta) in enumerate(train_loader):
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataloader.py", line 451, in iter
return _DataLoaderIter(self)
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataloader.py", line 247, in init
self._put_indices()
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataloader.py", line 295, in _put_indices
indices = next(self.sample_iter, None)
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/sampler.py", line 138, in iter
for idx in self.sampler:
File "/usr/local/lib/python3.6/dist-packages/torch/utils/data/sampler.py", line 51, in iter
return iter(torch.randperm(len(self.data_source)).tolist())
RuntimeError: randperm is only implemented for CPU

=======================================================
In my case, I need to install other versions of PyTorch to solve this issue.

Why is tr and tcl both represented by two channels? Is it ok to use one channel?

感谢你的开源，请问有没有实现作者的后处理，弯曲图像的透视变换，方便接入ocr。

HI @princewang1994, I found the logic to find snake head is a little different from the paper. But I have to say it is really difficult to handle. Compared with selecting the 2 edges with M nearest to -1. Here you use a proposal based method selecting all possible edges which has M < -0.7.

    if len(pts) > 4:
        e = np.concatenate([pts, pts[:3]])
        candidate = []
        for i in range(1, len(pts) + 1):
            v_prev = e[i] - e[i - 1]
            v_next = e[i + 2] - e[i + 1]
            if cos(v_prev, v_next) < -0.7:
                candidate.append((i % len(pts), (i + 1) % len(pts), norm2(e[i] - e[i + 1])))

The next refining step is somehow confusing.
if this logic is True:

candidate[0][0] == candidate[1][1] or candidate[0][1] == candidate[1][0]

you are visiting the same edge from 2 different direction, which seems impossible for me.

C:\w\1\s\windows\pytorch\aten\src\ATen/native/IndexingUtils.h:20: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead.

Warning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead. (expandTensors at C:/w/1/s/windows/pytorch/aten/src\ATen/native/IndexingUtils.h:20)

请问该怎么配合crnn使用，需要分割后对齐排列吗

Hello Author，I get a problem " loss=inf " when I train my data，it happend in several epoch, and the detail is here：
..........
(650 / 2000) - Loss: 1.3488 - tr_loss: 0.2789 - tcl_loss: 0.4046 - sin_loss: 0.1108 - cos_loss: 0.0985 - radii_loss: 0.4560
(700 / 2000) - Loss: 0.9112 - tr_loss: 0.2656 - tcl_loss: 0.3378 - sin_loss: 0.1111 - cos_loss: 0.0474 - radii_loss: 0.1492
(750 / 2000) - Loss: 1.3603 - tr_loss: 0.2677 - tcl_loss: 0.3537 - sin_loss: 0.0388 - cos_loss: 0.0105 - radii_loss: 0.6896
(800 / 2000) - Loss: 1.5277 - tr_loss: 0.2856 - tcl_loss: 0.3284 - sin_loss: 0.1668 - cos_loss: 0.0420 - radii_loss: 0.7048
/home/hj/smbshare/fffan/Detector/TextSnake/TextSnake.pytorch-master/util/misc.py:85: RuntimeWarning: invalid value encountered in double_scalars
return v[1] / l
/home/hj/smbshare/fffan/Detector/TextSnake/TextSnake.pytorch-master/util/misc.py:91: RuntimeWarning: invalid value encountered in double_scalars
return v[0] / l
(850 / 2000) - Loss: 2.1355 - tr_loss: 0.4698 - tcl_loss: 0.4413 - sin_loss: 0.1460 - cos_loss: 0.1237 - radii_loss: 0.9547
(900 / 2000) - Loss: 1.9340 - tr_loss: 0.4183 - tcl_loss: 0.3667 - sin_loss: 0.0760 - cos_loss: 0.0620 - radii_loss: 1.0110
(950 / 2000) - Loss: 1.2544 - tr_loss: 0.2943 - tcl_loss: 0.3723 - sin_loss: 0.0869 - cos_loss: 0.0938 - radii_loss: 0.4072
(1000 / 2000) - Loss: 1.2810 - tr_loss: 0.3134 - tcl_loss: 0.4176 - sin_loss: 0.1045 - cos_loss: 0.0622 - radii_loss: 0.3833
(1050 / 2000) - Loss: 2.0816 - tr_loss: 0.2053 - tcl_loss: 0.3180 - sin_loss: 0.0600 - cos_loss: 0.0517 - radii_loss: 1.4467
(1100 / 2000) - Loss: 1.7673 - tr_loss: 0.2696 - tcl_loss: 0.4861 - sin_loss: 0.1194 - cos_loss: 0.0957 - radii_loss: 0.7965
(1150 / 2000) - Loss: inf - tr_loss: 0.3392 - tcl_loss: 0.4790 - sin_loss: inf - cos_loss: 0.1815 - radii_loss: 0.1718

Traceback (most recent call last):
File "train_textsnake.py", line 239, in
main()
File "train_textsnake.py", line 224, in main
train(model, train_loader, criterion, scheduler, optimizer, epoch, logger)
File "train_textsnake.py", line 73, in train
criterion(output, tr_mask, tcl_mask, sin_map, cos_map, radius_map, train_mask)
File "/home/hj/.pyenv/versions/fffan-env/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in call
result = self.forward(*input, **kwargs)
File "/home/hj/smbshare/fffan/Detector/TextSnake/TextSnake.pytorch-master/network/loss.py", line 61, in forward
loss_tr = self.ohem(tr_pred, tr_mask.long(), train_mask.long())
File "/home/hj/smbshare/fffan/Detector/TextSnake/TextSnake.pytorch-master/network/loss.py", line 24, in ohem
loss_neg, _ = torch.topk(loss_neg, n_neg)
RuntimeError: merge_sort: failed to synchronize: an illegal memory access was encountered

And always get this error during training "util/misc.py:91: RuntimeWarning: invalid value encountered in double_scalars"
Can you solve those problems?
PS: My data is ICPR

Hi, thank you for your open source contributions.
I follow total_text.py write the data loading part of CTW 1500 and encountered the following problems
Traceback (most recent call last): File "CTW1500_debug.py", line 86, in <module> img, train_mask, tr_mask, tcl_mask, radius_map, sin_map, cos_map, meta = trainset[idx] File "CTW1500_debug.py", line 55, in __getitem__ return self.get_training_data(image, polygons, image_id=image_id, image_path=image_path) File "/work3/liuchao/TextSnake.pytorch/dataset/dataload.py", line 179, in get_training_data sideline1, sideline2, center_points, radius = polygon.disk_cover(n_disk=cfg.n_disk) File "/work3/liuchao/TextSnake.pytorch/dataset/dataload.py", line 53, in disk_cover inner_points2 = split_edge_seqence(self.points, self.e2, n_disk) File "/work3/liuchao/TextSnake.pytorch/util/misc.py", line 195, in split_edge_seqence while(cur_end > point_cumsum[cur_node + 1]): IndexError: index 1 is out of bounds for axis 0 with size 1
some of the cases have the above problems. I can understand the reason of the error, but I don't know how to correct it. Can you help me fix the bug or share your CTW1500 and ICDAR 2015 dataset file? Thank you very much.

Could you tell me what is the format of the dataset?

Could the ICDAR 2015 dataset work?

the data format like this 👍

x1 , y1 , x2, y2 , x3, y3 ,x4 ,y4 hello

thank you very much

I managed to finish the demo.py, but I got the .txt result, how should I convert it to the corresponding image result display?I am sorry to trouble you again.

Hi Dear Wang,
If I find out correctly,

1. Pretraining with SynthText section, train model using SynthText dataset.
2. In the Training section, we have 2 choises: a)training from scratch which train model with only TotalText dataset and b) training with pretrained model.

Is these correct?

My question is that: I have 100 images. How can I prepare them to training?
thanks

Hello,I found the Total-Text-dataset but I didn't see the download.sh file .

I have trained this model, but i only got F1 measure 58% nearly. do you have any ideas to improve the results. thank you. by the way, could you tell me your result.

train_mask = train_mask.view(-1) tr_mask = tr_mask.contiguous().view(-1)

After doing that, you use them by [train_mask * tr_mask]. Do you want to exclude the tough text?

Hi dear,

thank you for your code.
but my both train and validation loss don't decrease under 0.45 with 400 epochs. May you help me for more config such as learning rate and so on?

regards

Hi Dear. I want to use detector.merge_contours method to merge overlapped contours. but it gives an error "ValueError: too many values to unpack (expected 2)" at this line: cont_i, disk_i = all_contours[i]. thank you for your help. can you give insight that how does merge contour work? what are cont_i disk_i s? thanks

after executing demo.py, it detected the curved text in the image but didn't provide the flattened output as shown in figure 9. Can you help me with that?

Epoch: 0 : LR = 0.0001 Traceback (most recent call last): File "train_textsnake.py", line 238, in <module> main() File "train_textsnake.py", line 223, in main train(model, train_loader, criterion, scheduler, optimizer, epoch, logger) File "train_textsnake.py", line 63, in train for i, (img, train_mask, tr_mask, tcl_mask, radius_map, sin_map, cos_map, meta) in enumerate(train_loader): File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 345, in __next__ data = self._next_data() File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 856, in _next_data return self._process_data(data) File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 881, in _process_data data.reraise() File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/_utils.py", line 395, in reraise raise self.exc_type(msg) RuntimeError: Caught RuntimeError in DataLoader worker process 0. Original Traceback (most recent call last): File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/_utils/worker.py", line 178, in _worker_loop data = fetcher.fetch(index) File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/_utils/fetch.py", line 47, in fetch return self.collate_fn(data) File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 79, in default_collate return [default_collate(samples) for samples in transposed] File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 79, in <listcomp> return [default_collate(samples) for samples in transposed] File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 64, in default_collate return default_collate([torch.as_tensor(b) for b in batch]) File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 64, in <listcomp> return default_collate([torch.as_tensor(b) for b in batch]) File "/home/yt/anaconda3/envs/pytorch_zqs/lib/python3.6/site-packages/torch/cuda/__init__.py", line 148, in _lazy_init "Cannot re-initialize CUDA in forked subprocess. " + msg) RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method
Does anyone know how to solve it？ Thanks.

@princewang1994 Thanks for your excellent implementation.

I have tried vgg16 backbone. However, I suppose the resnet50 backbone is better than vgg16 in this case. How to support resnet50 backbone in the textnet.py?

Thanks

你好，请问中文文字检测的模型能分享一下吗

Hi, I download the dataset from here then ran python dataset/synth-text/make_list.py and I couldn't find image_list.txt.

Is it necessary to download the dataset from this link?

I have two questions about function mask_to_tcl in util/detection.py.

1. Why not change result.append(np.array([x_shift, y_shift, radii])) to result.append(np.array([x_c, y_c, radii]))? I tried and F score raised from 0.657 to 0.663, and it makes more sense.
2. What are x_shift_pos and x_shift_neg's purpose?

After transform, polygons will change. Bottom and side line still correct for disk cover?

Could you please point me to the code of the transformation done for figure-9 of the paper ?

Thanks for your great work and codes.

I followed your commands:

1. I used pre-train phase.
2. used total-text dataset to improve training model.
3. test on total-text database to evaluate model.

now I want to evaluate model on my sample images. I run demo.py but it need mat file of my sample images, while I dont want to give .mat file. I want to test for images in which i don't have their gt files.

regards,John

i am getting below error when i try to test pretrained model on my image

=============End=============
Loading from ./save/pretrained/textsnake_vgg_180.pth
Start testing TextSnake.
Traceback (most recent call last):
File "demo.py", line 110, in
main()
File "demo.py", line 92, in main
inference(detector, test_loader, output_dir)
File "demo.py", line 46, in inference
contours, output = detector.detect(image)
File "/home/ec2-user/vinayak/TextSnake.pytorch/util/detection.py", line 239, in detect
tr_pred = output[0, 0:2].softmax(dim=0).data.cpu().numpy()
AttributeError: 'Tensor' object has no attribute 'softmax'

I over-fit on a small sample of TotalText, until I got a loss of about 0.1 in 6 images, but the demo file with the trained checkpoint doesn't show any contours.

does the total_text.py can only deal with curved texts ?

I tried put minimum enclosing rectangle text instances to recognition model but the results very poor. I saw Figure 9 in the article but didn't know how to do it. Can you suggest some algorithm for do it

Hello,

torch.load does not manage to load correctly textsnake_vgg_0.pth (downloaded with google drive) using google colab which is linked to my drive. Here is the command I ran:

! python3.6 train_textsnake.py example --viz --batch_size 8 --resume /content/gdrive/My\ Drive/Colab\ Notebooks/TextSnake.pytorch/save/synthtext_pretrain/textsnake_vgg_0.pth

Here is the error I get (straight after it printed all the configs):

Loading from /content/gdrive/My Drive/Colab Notebooks/TextSnake.pytorch/save/synthtext_pretrain/textsnake_vgg_0.pth
Traceback (most recent call last):
File "train_textsnake.py", line 238, in
main()
File "train_textsnake.py", line 210, in main
load_model(model, cfg.resume)
File "train_textsnake.py", line 46, in load_model
state_dict = torch.load(model_path)['state_dict']
File "/usr/local/lib/python3.6/dist-packages/torch/serialization.py", line 303, in load
return _load(f, map_location, pickle_module)
File "/usr/local/lib/python3.6/dist-packages/torch/serialization.py", line 454, in _load
return legacy_load(f)
File "/usr/local/lib/python3.6/dist-packages/torch/serialization.py", line 380, in legacy_load
with closing(tarfile.open(fileobj=f, mode='r:', format=tarfile.PAX_FORMAT)) as tar,
File "/usr/lib/python3.6/tarfile.py", line 1589, in open
return func(name, filemode, fileobj, **kwargs)
File "/usr/lib/python3.6/tarfile.py", line 1619, in taropen
return cls(name, mode, fileobj, **kwargs)
File "/usr/lib/python3.6/tarfile.py", line 1482, in init
self.firstmember = self.next()
File "/usr/lib/python3.6/tarfile.py", line 2297, in next
tarinfo = self.tarinfo.fromtarfile(self)
File "/usr/lib/python3.6/tarfile.py", line 1092, in fromtarfile
buf = tarfile.fileobj.read(BLOCKSIZE)
OSError: [Errno 5] Input/output error

I assume part of the error is that pytorch is trying to open as a tar file although it is not.

I would really appreciate your help.

Best Wishes.

The point removal of in TextInstance removes all qualified points independently. Which is not working well in some cases.

before [[  15    0]
 [1216    0]
 [1219    0]
 [  13   16]]

after [[15  0]
 [13 16]]

The image is from the latest ArT dataset.

I do not understand loss_radii = F.smooth_l1_loss(radii_pred[tcl_mask] / radii_map[tcl_mask], ones), what do radii_pred[tcl_mask] and radii_map[tcl_mask] mean? Can someone help me?

First , Wish you a Happy New Year.

I have some problem ,maybe you can help me . Thakn you very much.

I trained the model with my datasets . Now the detector results is some ploygons which have three
formats (center point , tangent line,text region). How can I use the results to get the rectangle or square?

Thank you very much.

thanks for your sharing, however, while i download total-text files from url:
https://docs.google.com/uc?export=download&id=1bC68CzsSVTusZVvOkk7imSZSbgD1MqK2
and unzip it. and run train_textsnake.py, the errors occur:
FileNotFoundError: [Errno 2] No such file or directory: 'data/total-text/gt/Train/poly_gt_img697.mat'

运行你的训练脚本，结果报下面这个错误：
Traceback (most recent call last):
File "train_textsnake.py", line 239, in
main()
File "train_textsnake.py", line 224, in main
train(model, train_loader, criterion, scheduler, optimizer, epoch, logger)
File "train_textsnake.py", line 63, in train
for i, (img, train_mask, tr_mask, tcl_mask, radius_map, sin_map, cos_map, meta) in enumerate(train_loader):
File "/home/hj/anaconda3/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 582, in next
return self._process_next_batch(batch)
File "/home/hj/anaconda3/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 608, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
RuntimeError: Traceback (most recent call last):
File "/home/hj/anaconda3/lib/python3.5/site-packages/torch/utils/data/_utils/worker.py", line 99, in _worker_loop
samples = collate_fn([dataset[i] for i in batch_indices])
File "/home/hj/anaconda3/lib/python3.5/site-packages/torch/utils/data/_utils/collate.py", line 68, in default_collate
return [default_collate(samples) for samples in transposed]
File "/home/hj/anaconda3/lib/python3.5/site-packages/torch/utils/data/_utils/collate.py", line 68, in
return [default_collate(samples) for samples in transposed]
File "/home/hj/anaconda3/lib/python3.5/site-packages/torch/utils/data/_utils/collate.py", line 63, in default_collate
return {key: default_collate([d[key] for d in batch]) for key in batch[0]}
File "/home/hj/anaconda3/lib/python3.5/site-packages/torch/utils/data/_utils/collate.py", line 63, in
return {key: default_collate([d[key] for d in batch]) for key in batch[0]}
File "/home/hj/anaconda3/lib/python3.5/site-packages/torch/utils/data/_utils/collate.py", line 59, in default_collate
return torch.tensor(batch)
RuntimeError: CUDA error: initialization error

相关配置为：
==========Options============
input_channel: 1
log_freq: 100
num_workers: 8
optim: SGD
max_annotation: 200
gamma: 0.1
val_freq: 100
mgpu: False
viz: False
lr: 0.0001
display_freq: 50
save_freq: 10
loss: CrossEntropyLoss
lr_adjust: fix
exp_name: ICPR
pretrain: False
viz_freq: 50
cuda: True
rescale: 255.0
stepvalues: []
weight_decay: 0.0
momentum: 0.9
max_points: 20
resume: None
start_iter: 0
input_size: 512
batch_size: 4
img_root: /home/hj/smbshare/fffan/Data/ICPR_text_train_20180313/train_image_8000
means: [0.485, 0.456, 0.406]
tr_thresh: 0.6
start_epoch: 0
use_hard: True
save_dir: ./save/
max_epoch: 200
checkepoch: -1
log_dir: ./logs/
net: vgg
verbose: True
output_dir: output
device: cuda
post_process_merge: False
dataset: synth-text
post_process_expand: 0.3
n_disk: 15
tcl_thresh: 0.4
stds: [0.229, 0.224, 0.225]
vis_dir: ./vis/

您好，这个是说训练集的坐标有问题吗？
@princewang1994

Hello,

When trying to run
CUDA_VISIBLE_DEVICES=0 python train.py train_testy
I get the following error:

Traceback (most recent call last):
  File "train.py", line 157, in <module>
    main()
  File "train.py", line 143, in main
    train(model, train_loader, criterion, scheduler, optimizer, epoch)
  File "train.py", line 45, in train
    for i, (img, train_mask, tr_mask, tcl_mask, radius_map, sin_map, cos_map, meta) in enumerate(train_loader):
  File "/home/anaconda3/envs/text_snake_env/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 513, in __next__
    return self._process_next_batch(batch)
  File "/home/anaconda3/envs/text_snake_env/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 534, in _process_next_batch
    raise batch.exc_type(batch.exc_msg)
RuntimeError: Traceback (most recent call last):
  File "/home/anaconda3/envs/text_snake_env/lib/python3.6/site-packages/torch/utils/data/_utils/worker.py", line 99, in _worker_loop
    samples = collate_fn([dataset[i] for i in batch_indices])
  File "/home/anaconda3/envs/text_snake_env/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 66, in default_collate
    return [default_collate(samples) for samples in transposed]
  File "/home/anaconda3/envs/text_snake_env/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 66, in <listcomp>
    return [default_collate(samples) for samples in transposed]
  File "/home/anaconda3/envs/text_snake_env/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 63, in default_collate
    return {key: default_collate([d[key] for d in batch]) for key in batch[0]}
  File "/home/anaconda3/envs/text_snake_env/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 63, in <dictcomp>
    return {key: default_collate([d[key] for d in batch]) for key in batch[0]}
  File "/home/anaconda3/envs/text_snake_env/lib/python3.6/site-packages/torch/utils/data/_utils/collate.py", line 59, in default_collate
    return torch.tensor(batch)
RuntimeError: CUDA error: initialization error

It looks like it has something to do with enumerator(train_loader)

Trying to run pure inference part, Downgraded pytorch version to 1.5 (torch==1.5.0+cu101 torchvision==0.6.0+cu101), then Set CUDA_VISIBLE_DEVICES=0 python demo.py $EXPNAME --checkepoch 180 --img_root "path to image " , but getting following error in dataset/dataload.py

遇到一些问题想请教一下作者。首先是bash download.sh运行失败。索性自己去github地址上面下下来了total-text数据集。然而github上面的totaltext 其gt使用txt标记。于是首先吧txt改为了csv文件，然后修改total-text.py中的parse_mat()函数为如下代码：
'''
def parse_csv(self, csv_path):
with open(csv_path,'rt', encoding='UTF-8') as row_data:
readers = csv.reader(row_data, delimiter=',')
data = list(readers)

        polygons = []
        for cell in data:
            x = [int(num) for num in re.findall('\d+', str(cell[0].strip()))]
            y = [int(num) for num in re.findall('\d+', str(cell[1].strip()))]

            text = re.findall('\'([^>]+?)\'', str(cell[3]))[0]
            ori = re.findall('\'([^>]+?)\'', str(cell[2]))[0]

            text = text if len(text)>0 else '#'
            ori = ori if len(ori)>0 else 'c'

            if len(x) < 4:  # too few points
                continue
            pts = np.stack([x, y]).T.astype(np.int32)
            polygons.append(TextInstance(pts, ori, text))

    return polygons

'''

最后这是我的运行结果：
'''
aceback (most recent call last):
File "eval_textsnake.py", line 123, in
main()
File "eval_textsnake.py", line 102, in main
inference(detector, test_loader, output_dir)
File "eval_textsnake.py", line 48, in inference
contours, output = detector.detect(image)
File "/root/fsy_SceneTextRec/Docker-pytorch0.4.1/TextSnake.pytorch/util/detection.py", line 237, in detect
output = self.model(image)
File "/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in call
result = self.forward(*input, **kwargs)
File "/root/fsy_SceneTextRec/Docker-pytorch0.4.1/TextSnake.pytorch/network/textnet.py", line 47, in forward
C1, C2, C3, C4, C5 = self.backbone(x)
File "/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in call
result = self.forward(*input, **kwargs)
File "/root/fsy_SceneTextRec/Docker-pytorch0.4.1/TextSnake.pytorch/network/vgg.py", line 92, in forward
C1 = self.stage1(x)
File "/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in call
result = self.forward(*input, **kwargs)
File "/opt/conda/lib/python3.6/site-packages/torch/nn/modules/container.py", line 91, in forward
input = module(input)
File "/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in call
result = self.forward(*input, **kwargs)
File "/opt/conda/lib/python3.6/site-packages/torch/nn/modules/conv.py", line 301, in forward
self.padding, self.dilation, self.groups)
RuntimeError: cuda runtime error (11) : invalid argument at /opt/conda/conda-bld/pytorch_1532579805626/work/aten/src/THC/THCGeneral.cpp:663
'''
运行环境是RTX2070，pytorch0.4.1的官方docker，cuda9.0

could you share the trained model?

thank you very much

请问为什么tr和tcl都用两通道表示？各用一通道表示不就可以了吗

您好，您有测过复现的精度能达到或者接近作者的精度吗

Loading from ./save/example/textsnake_vgg_10.pth
Start testing TextSnake.
detect 0 / 300 images: img650.jpg.
Traceback (most recent call last):
  File "demo.py", line 133, in <module>
    main()
  File "demo.py", line 119, in main
    inference(model, detector, test_loader)
  File "demo.py", line 77, in inference
    batch_result = detector.detect(tr_pred, tcl_pred, sin_pred, cos_pred, radii_pred)  # (n_tcl, 3)
  File "/home/crypa/TextSnake.pytorch/util/detection.py", line 204, in detect
    detect_result = self.build_tcl(tcl, sin_pred, cos_pred, radii_pred)
  File "/home/crypa/TextSnake.pytorch/util/detection.py", line 164, in build_tcl
    _, conts, _ = cv2.findContours(mask.astype(np.uint8), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
ValueError: not enough values to unpack (expected 3, got 2)

Due to time, I didn't reach 200epoch according to the training requirements.
I tried to test the code with 10epoch.
so run CUDA_VISIBLE_DEVICES=0 python demo.py --checkepoch 10 example
But got the above error.so sorry,I am a newbie,could u help me?