Pretrained Models

[avinabsaha]

@JasonBoy1 can you release pre-trained models?

[WenjiaWang0312]

The models were deleted by me... Maybe I should train it recently.

[avinabsaha]

@JasonBoy1 could you share models when they are available?

[WenjiaWang0312]

@JasonBoy1 could you share models when they are available?

Sure.

[anuda]

Just wondering if you had chance to build the model? Thanks in advance!

[WenjiaWang0312]

Just wondering if you had chance to build the model? Thanks in advance!

Sorry, too busy recently.

[yustiks]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing

I hope, it is helpful.

[dkaliroff]

@yustiks thanks for sharing. Did you test the model? does it achieve similar results as in the paper ? @JasonBoy1 could you please share your pretrained model if you have them ? Thanks

[avinabsaha]

@yustiks please let us know about the performance of your trained model in comparison to the results mentioned in the paper.

[yustiks]

@dkaliroff @avinabsaha currently, I can't evaluate the model due to error mentioned here: https://github.com/JasonBoy1/TextZoom/issues/28

but you can try to evaluate by yourself.

Visually, the results are good.

[yustiks]

@dkaliroff @avinabsaha for example, aster test prediction is: {'accuracy': {'easy': 0.7332}, 'psnr_avg': 23.423454, 'ssim_avg': 0.866824, 'fps': 18.50144357823526} {'accuracy': {'medium': 0.562}, 'psnr_avg': 18.642635, 'ssim_avg': 0.659524, 'fps': 16.28411871321792} {'accuracy': {'hard': 0.3917}, 'psnr_avg': 19.63501, 'ssim_avg': 0.731286, 'fps': 17.888223484995095}

which is comparable with the paper results Real: 'easy': 75.1% 'medium': 56.3% 'hard': 40.1% 'average': 58.3%

[anuda]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing

I hope, it is helpful.

Thank you! i will give it a try

[mpolvere96]

I have tried but I have CUDA issues. What environment did you use?

[mpolvere96]

I have this error Traceback (most recent call last): File “main.py”, line 46, in main(config, args) File “main.py”, line 19, in main Mission.train() File “/opt/notebooks/TextZoom/src/interfaces/super_resolution.py”, line 33, in train model_dict = self.generator_init() File “/opt/notebooks/TextZoom/src/interfaces/base.py”, line 149, in generator_init model = model.to(self.device) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 381, in to return self._apply(convert) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 187, in _apply module._apply(fn) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 187, in _apply module._apply(fn) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 187, in _apply module._apply(fn) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/rnn.py”, line 117, in _apply self.flatten_parameters() File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/rnn.py”, line 113, in flatten_parameters self.batch_first, bool(self.bidirectional)) RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED

[yustiks]

I have this error Traceback (most recent call last): File “main.py”, line 46, in main(config, args) File “main.py”, line 19, in main Mission.train() File “/opt/notebooks/TextZoom/src/interfaces/super_resolution.py”, line 33, in train model_dict = self.generator_init() File “/opt/notebooks/TextZoom/src/interfaces/base.py”, line 149, in generator_init model = model.to(self.device) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 381, in to return self._apply(convert) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 187, in _apply module._apply(fn) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 187, in _apply module._apply(fn) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 187, in _apply module._apply(fn) File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/rnn.py”, line 117, in _apply self.flatten_parameters() File “/root/.conda/envs/textZoom/lib/python3.6/site-packages/torch/nn/modules/rnn.py”, line 113, in flatten_parameters self.batch_first, bool(self.bidirectional)) RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED

Seems, there is some problem with CUDA drivers.

[mpolvere96]

Thank you! Which version of cuda and torch did you use?

[JerryLeolfl]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing

I hope, it is helpful.

Hi, bro. Thanks for your model, bur the url is not available now. Could you please provide a new address to get it.

[JerryLeolfl]

@yustiks please let us know about the performance of your trained model in comparison to the results mentioned in the paper.

@avinabsaha

Just wondering if you had chance to build the model? Thanks in advance!

Sorry, too busy recently.

@JasonBoy1 Thanks for your great work. Is the pretrained model available now?

[yustiks]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing I hope, it is helpful.

Hi, bro. Thanks for your model, bur the url is not available now. Could you please provide a new address to get it.

should work now

[xianglei96]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing

I hope, it is helpful.

I try to test in textzoom but the result is all noise.

[yustiks]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing I hope, it is helpful.

I try to test in textzoom but the result is all noise.

In this picture, what is the input, and what is the output?

[xianglei96]

邮件已收到，我会尽快回复~

[xianglei96]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing I hope, it is helpful.

I try to test in textzoom but the result is all noise.

In this picture, what is the input, and what is the output?

[yustiks]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing I hope, it is helpful.

I try to test in textzoom but the result is all noise.

In this picture, what is the input, and what is the output?

Interesting. How did you use weights to run the model?

[xianglei96]

https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing

Use the model provided by you, the dataset and code provided by the author.

[sfxjh]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing I hope, it is helpful.

Hi, bro. Thanks for your model, bur the url is not available now. Could you please provide a new address to get it.

should work now

I had a weird problem. When I try to import your model, I see a problem with the size of the modules:

size mismatch for module.block1.0.weight: copying a param with shape torch.Size([64, 4, 9, 9]) from checkpoint, the shape in current model is torch.Size([64, 3, 9, 9]). size mismatch for module.block8.1.weight: copying a param with shape torch.Size([4, 64, 9, 9]) from checkpoint, the shape in current model is torch.Size([3, 64, 9, 9]). size mismatch for module.block8.1.bias: copying a param with shape torch.Size([4]) from checkpoint, the shape in current model is torch.Size([3])

Have you ever encountered such a problem?

[xianglei96]

邮件已收到，我会尽快回复~

[yustiks]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing I hope, it is helpful.

Hi, bro. Thanks for your model, bur the url is not available now. Could you please provide a new address to get it.

should work now

I had a weird problem. When I try to import your model, I see a problem with the size of the modules:

size mismatch for module.block1.0.weight: copying a param with shape torch.Size([64, 4, 9, 9]) from checkpoint, the shape in current model is torch.Size([64, 3, 9, 9]). size mismatch for module.block8.1.weight: copying a param with shape torch.Size([4, 64, 9, 9]) from checkpoint, the shape in current model is torch.Size([3, 64, 9, 9]). size mismatch for module.block8.1.bias: copying a param with shape torch.Size([4]) from checkpoint, the shape in current model is torch.Size([3])

Have you ever encountered such a problem?

From what I can see, your input size is just one image, but instead, it should be the list of images. Try to create a new array, and add the image as an element to this array, after that, run a model.

[Everythingismetaphor]

I uploaded the best model weights after 42000 epochs of training on TextZoom data: https://drive.google.com/file/d/1j-g17V5kBmqS8giWNZuHe2d_GefAepCs/view?usp=sharing

I hope, it is helpful. What is your learning rate? If I set batch size 16,what should the learning rate be set to?

[xianglei96]

邮件已收到，我会尽快回复~