-
Hello!
@ChWick
As described the architecture of the network described in "/calamari_ocr/ocr/backends/tensorflow_backend/tensorflow_model.py"
I try to use one of the existant pretrained models :
…
-
I try to modify tedlium2/align1 CTCsegment example to another example like egs/aishell
but report error as following
I have no idea to solve this index error.
```2020-09-28 01:37:54,391 (asr_ali…
-
## 🐛 Bug
When I use a single GPU, all parameters in the model get non-zero gradients;
but after wrapping my model with nn.DataParallel, the gradients of some part (RNN in my case) of the model bec…
-
Hi there !
Thank you for the toolkit 👍
First, I'm wondering if it would be possible to not perform seq2seq training for ASR. Let's say I have a BLSTM model, and I want to directly train with C…
-
I am looking for a suitable way for automating the _**Captcha Recognition**_.
**Things I have already tried:**
I have tried using pytesseract but it didn't work. I had pre-processed the captcha im…
-
I am trying ASR task with Mozilla common voice dataset using a pure attention model (mtalpha=0, ctc_weight=0). However, while calculating validation values, the cer and loss values are significantly h…
-
I'm now revisiting CTC, attention, and hybrid CTC/attention, and found that chainer training is very slow compared with pytorch training even when we only use CTC. This is strange for me because our C…
-
It seems that we can easily use multiple GPUs (data parallel) with chainer trainer. I want someone to consider this extension. Several tasks (e.g., Librispeech) take around one week, even with the pyt…
-
I have a little question about this part below. Does this mean you slice it along first axis, which means you slice it along batch-size dimension? But according to the paper,shouldn't it be sliced alo…
-
Hi, I would like to use this code to extract features using CNN, I'm asking if I can use a sliding window width more than 1 pixel .
My goal is to extract a set of features based on CNN and to train a…