Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/edward-zhu/umaru

An OCR-system based on torch using the technique of LSTM/GRU-RNN, CTC and referred to the works of rnnlib and clstm.
https://github.com/edward-zhu/umaru

Last synced: 6 days ago
JSON representation

An OCR-system based on torch using the technique of LSTM/GRU-RNN, CTC and referred to the works of rnnlib and clstm.

Host: GitHub
URL: https://github.com/edward-zhu/umaru
Owner: edward-zhu
License: other
Created: 2015-08-03T06:54:52.000Z (over 9 years ago)
Default Branch: master
Last Pushed: 2015-10-27T07:27:21.000Z (about 9 years ago)
Last Synced: 2024-08-03T04:05:44.094Z (4 months ago)
Language: Lua
Homepage:
Size: 1.49 MB
Stars: 66
Watchers: 4
Forks: 20
Open Issues: 2
Metadata Files:
- Readme: README.md
- License: LICENSE

Awesome Lists containing this project

awesome-ocr - An OCR-system based on Torch using the technique of LSTM/GRU-RNN, CTC and referred to the works of rnnlib and clstm.

README

# umaru
An OCR-system based on torch using the technique of LSTM/GRU-RNN, CTC and referred to the works of rnnlib and clstm.

## Notice

This work is now completely UNSTABLE, EXPERIMENTAL and UNDER DEVELOPMENT.

## Dependencies

- [torch](https://github.com/torch/torch7) (and following packages)
- image
- nn/cunn
- optim
- [rnn](https://github.com/Element-Research/rnn)
- [json](https://github.com/clementfarabet/lua---json)
- [utf8](https://github.com/clementfarabet/lua-utf8)
- [torchRBM](https://github.com/nhammerla/torchRBM)

## Build

```sh
$ ./build.sh
```

## Usage

### General

- You could modify the settings in the `main.lua` directly and execute `th main.lua`, the input format is clstm-like (`.png` and `.gt.txt` pair) and you should put all input file path in a text file.
- or if you prefer to use a JSON-format configuration file, you could follow the example below, and run:

```sh
$ th main.lua -setting [setting file]
```

### Run Folder

There would be a folder created in the `experments` folder for every experiment. You could check out the log, settings and saved models there.

## Example Configuration File

descriptions for each option could be found in `main.lua`.

```js
{
"project_name": "uy_rbm_noised",
"raw_input": false,
"hidden_size": 200,
"nthread": 3,
"clamp_size": 1,
"ctc_lua": false,
"recurrent_unit": "gru",
"test_every": 2000,
"omp_threads": 1,
"show_every": 10,
"testing_list_file": "wwr.txt",
"input_size": 48,
"testing_ratio": 1,
"max_param_norm": false,
"training_list_file": "full-train.txt",
"feature_size": 240,
"momentum": 0.9,
"dropout_rate": 0.5,
"max_iter": 10000000000,
"save_every": 10000,
"learning_rate": 0.0001,
"stride": 5,
"gpu": false,
"rbm_network_file": "rbm/wwr.rbm",
"windows_size": 10
}
```

## LICENSE

BSD 3-Clause License

## References

* [Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling](http://arxiv.org/abs/1412.3555)
* [Connectionist Temporal Classification: Labelling Unsegmented Sequence Data with Recurrent Neural Networks](ftp://ftp.idsia.ch/pub/juergen/icml2006.pdf)
* [RNNLIB: Connectionist Temporal Classification and Transcription Layer](http://wantee.github.io/blog/2015/02/08/rnnlib-connectionist-temporal-classification-and-transcription-layer/)
* [rnnlib](http://sourceforge.net/p/rnnl/wiki/Home/)
* [clstm](https://github.com/tmbdev/clstm)