-
Notifications
You must be signed in to change notification settings - Fork 27
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
adding some more edits to code to handle synth data
- Loading branch information
1 parent
ad2e125
commit 85cef8a
Showing
17 changed files
with
453 additions
and
20 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,2 +1,53 @@ | ||
| # Adapting-OCR | ||
| Pytorch implementation of our [paper](http://cdn.iiit.ac.in/cdn/cvit.iiit.ac.in/images/ConferencePapers/2020/AdaptingOCR_Deepayan_DAS2020_final.pdf) | ||
| Pytorch implementation of our [Adapting OCR with limited labels](http://cdn.iiit.ac.in/cdn/cvit.iiit.ac.in/images/ConferencePapers/2020/AdaptingOCR_Deepayan_DAS2020_final.pdf) | ||
|
|
||
|  | ||
|
|
||
| ## Dependency | ||
|
|
||
| * This work was tested with PyTorch 1.2.0, CUDA 9.0, python 3.6 and Ubuntu 16.04. | ||
| * requirements can be found in the file. | ||
| * command to create environment from the file is `conda create -n pytorch1.4 --file env.txt` | ||
| * To activate the environment use: `source activate pytorch1.4` | ||
|
|
||
| ## Training | ||
|
|
||
| * Supervised training | ||
|
|
||
| `python -m train --name exp1 --path path/to/data ` | ||
|
|
||
| * Main arguments | ||
| * `--name`: creates a directory where checkpoints will be stored | ||
| * `--path`: path to dataset. | ||
| * `--imgdir`: dir name of dataset | ||
|
|
||
|
|
||
| * Semi-supervised training | ||
|
|
||
| `python -m train_semi_supervised --name exp1 --path path --source_dir src_dirname --target_dir tgt_dirname --schedule --noise --alpha=1` | ||
|
|
||
| * Main arguments | ||
| * `--name`: creates a directory where checkpoints will be stored | ||
| * `--path`: path to datasets | ||
| * `--source_dir`: labelled data directory on which ocr was trained | ||
| * `--target_dir`: unlabeled data directory on which we want to adapt ocr | ||
| * `--percent`: percentage of unlabeled data to include in self-training | ||
| * `--schedule`: will include STLR scheduler while training | ||
| * `--train_on_pred`: will treat top-predictions as targets | ||
| * `--noise`: will add gaussian noise to images while training | ||
| * `--alpha`: set to 1 to include the mixup criterion | ||
| * `--combine_scoring`: will also take into account the scores outputted by a language model | ||
|
|
||
| **Note**: `--combine_scoring` works only with line images not word images | ||
|
|
||
| * Data | ||
| * Use [trdg](https://github.com/Belval/TextRecognitionDataGenerator) to generate synthetic data. The script for data generation is included `scrips/generate_data.sh`. | ||
| * Download two different fonts and keep the data pertaining to each font in source and target dirs. | ||
| * Use one of the fonts to train data from scratch in a supervised manner. | ||
| * Then finetune the trained model on target data using semi-supervised learning | ||
| * A sample lexicon is provided in `words.txt`. Download different lexicon as per need. | ||
|
|
||
|
|
||
|
|
||
|
|
||
|
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,156 @@ | ||
| # This file may be used to create an environment using: | ||
| # $ conda create --name <env> --file <this file> | ||
| # platform: linux-64 | ||
| _libgcc_mutex=0.1=main | ||
| _tflow_select=2.3.0=mkl | ||
| absl-py=0.9.0=py37hc8dfbb8_1 | ||
| astor=0.7.1=py_0 | ||
| blas=2.12=openblas | ||
| brotlipy=0.7.0=py37h8f50634_1000 | ||
| bzip2=1.0.8=h516909a_0 | ||
| c-ares=1.15.0=h516909a_1001 | ||
| ca-certificates=2019.9.11=hecc5488_0 | ||
| cairo=1.16.0=hfb77d84_1002 | ||
| certifi=2019.9.11=py37_0 | ||
| cffi=1.12.3=py37h2e261b9_0 | ||
| chardet=3.0.4=py37hc8dfbb8_1006 | ||
| cloudpickle=1.3.0=py_0 | ||
| cryptography=2.8=py37h72c5cf5_1 | ||
| cudatoolkit=10.0.130=0 | ||
| cycler=0.10.0=py_1 | ||
| cytoolz=0.10.1=py37h7b6447c_0 | ||
| dask-core=2.10.1=py_0 | ||
| dbus=1.13.6=he372182_0 | ||
| decorator=4.4.1=py_0 | ||
| expat=2.2.5=he1b5a44_1003 | ||
| ffmpeg=4.1.3=h167e202_0 | ||
| fire=0.2.1=py_0 | ||
| fontconfig=2.13.1=h86ecdb6_1001 | ||
| freetype=2.10.0=he983fc9_1 | ||
| gast=0.2.2=py_0 | ||
| gettext=0.19.8.1=hc5be6a0_1002 | ||
| giflib=5.1.9=h516909a_0 | ||
| glib=2.58.3=h6f030ca_1002 | ||
| gmp=6.1.2=hf484d3e_1000 | ||
| gnutls=3.6.5=hd3a4fd2_1002 | ||
| graphite2=1.3.13=hf484d3e_1000 | ||
| grpcio=1.23.0=py37he9ae1f9_0 | ||
| gst-plugins-base=1.14.5=h0935bb2_0 | ||
| gstreamer=1.14.5=h36ae1b5_0 | ||
| h5py=2.10.0=nompi_py37h513d04c_102 | ||
| harfbuzz=2.4.0=h9f30f68_3 | ||
| hdf5=1.10.5=nompi_h3c11f04_1103 | ||
| icu=64.2=he1b5a44_1 | ||
| idna=2.9=py_1 | ||
| intel-openmp=2019.4=243 | ||
| jasper=1.900.1=h07fcdf6_1006 | ||
| jpeg=9c=h14c3975_1001 | ||
| kiwisolver=1.1.0=py37hc9558a2_0 | ||
| krb5=1.16.3=h05b26f9_1001 | ||
| lame=3.100=h14c3975_1001 | ||
| libblas=3.8.0=12_openblas | ||
| libcblas=3.8.0=12_openblas | ||
| libclang=8.0.1=hc9558a2_0 | ||
| libcurl=7.65.3=hda55be3_0 | ||
| libedit=3.1.20181209=hc058e9b_0 | ||
| libffi=3.2.1=hd88cf55_4 | ||
| libgcc-ng=9.1.0=hdf63c60_0 | ||
| libgfortran-ng=7.3.0=hdf63c60_0 | ||
| libiconv=1.15=h516909a_1005 | ||
| liblapack=3.8.0=12_openblas | ||
| liblapacke=3.8.0=12_openblas | ||
| libllvm8=8.0.1=hc9558a2_0 | ||
| libopenblas=0.3.7=h6e990d7_1 | ||
| libpng=1.6.37=hed695b0_0 | ||
| libprotobuf=3.8.0=h8b12597_0 | ||
| libsodium=1.0.17=h516909a_0 | ||
| libssh2=1.8.2=h22169c7_2 | ||
| libstdcxx-ng=9.1.0=hdf63c60_0 | ||
| libtiff=4.0.10=h57b8799_1003 | ||
| libuuid=2.32.1=h14c3975_1000 | ||
| libwebp=1.0.2=h576950b_1 | ||
| libxcb=1.13=h14c3975_1002 | ||
| libxkbcommon=0.8.4=h516909a_0 | ||
| libxml2=2.9.9=hea5a465_1 | ||
| libxslt=1.1.32=h31b3aaa_1004 | ||
| lxml=4.4.1=py37h7ec2d77_0 | ||
| lz4-c=1.8.3=he1b5a44_1001 | ||
| markdown=3.2.1=py_0 | ||
| matplotlib=3.1.1=py37_1 | ||
| matplotlib-base=3.1.1=py37he7580a8_1 | ||
| mkl=2019.4=243 | ||
| mkl-service=2.0.2=py37h7b6447c_0 | ||
| mkl_fft=1.0.14=py37h516909a_1 | ||
| mkl_random=1.0.4=py37hf2d7682_0 | ||
| mock=3.0.5=py37hc8dfbb8_1 | ||
| ncurses=6.1=he6710b0_1 | ||
| nettle=3.4.1=h1bed415_1002 | ||
| networkx=2.4=py_0 | ||
| ninja=1.9.0=py37hfd86e86_0 | ||
| nspr=4.20=hf484d3e_1000 | ||
| nss=3.45=he751ad9_0 | ||
| numpy=1.16.4=py37h99e49ec_0 | ||
| numpy-base=1.16.4=py37h2f8d375_0 | ||
| olefile=0.46=py37_0 | ||
| opencv=4.1.1=py37ha799480_1 | ||
| openh264=1.8.0=hdbcaa40_1000 | ||
| openssl=1.1.1c=h516909a_0 | ||
| opt_einsum=3.2.0=py_0 | ||
| pcre=8.41=hf484d3e_1003 | ||
| pillow=6.2.1=py37h34e0f95_0 | ||
| pip=19.2.2=py37_0 | ||
| pixman=0.38.0=h516909a_1003 | ||
| protobuf=3.8.0=py37he1b5a44_2 | ||
| pthread-stubs=0.4=h14c3975_1001 | ||
| pycparser=2.19=py37_0 | ||
| pyopenssl=19.1.0=py_1 | ||
| pyparsing=2.4.2=py_0 | ||
| pyqt=5.9.2=py37hcca6a23_4 | ||
| pysocks=1.7.1=py37hc8dfbb8_1 | ||
| python=3.7.4=h265db76_1 | ||
| python-dateutil=2.8.0=py_0 | ||
| python-levenshtein=0.12.0=py37h516909a_1001 | ||
| python_abi=3.7=1_cp37m | ||
| pytorch=1.4.0=py3.7_cuda10.0.130_cudnn7.6.3_0 | ||
| pywavelets=1.1.1=py37h7b6447c_0 | ||
| pyzmq=19.0.1=py37hac76be4_0 | ||
| qt=5.9.7=h0c104cb_3 | ||
| readline=7.0=h7b6447c_5 | ||
| requests=2.23.0=pyh8c360ce_2 | ||
| scipy=1.4.1=py37habc2bb6_0 | ||
| setuptools=41.0.1=py37_0 | ||
| sip=4.19.8=py37hf484d3e_1000 | ||
| six=1.12.0=py37_0 | ||
| speechrecognition=3.6.3=py37_1000 | ||
| sqlite=3.29.0=h7b6447c_0 | ||
| tensorflow-base=1.15.0=mkl_py37he1670d9_0 | ||
| tensorflow-estimator=1.15.1=pyh2649769_0 | ||
| termcolor=1.1.0=py_2 | ||
| tk=8.6.9=hed695b0_1003 | ||
| toolz=0.10.0=py_0 | ||
| torchfile=0.1.0=py_0 | ||
| torchvision=0.5.0=py37_cu100 | ||
| tornado=6.0.3=py37h516909a_0 | ||
| tqdm=4.35.0=py_0 | ||
| urllib3=1.25.9=py_0 | ||
| visdom=0.1.8.9=0 | ||
| websocket-client=0.57.0=py37hc8dfbb8_1 | ||
| werkzeug=0.16.1=py_0 | ||
| wheel=0.33.4=py37_0 | ||
| wrapt=1.12.1=py37h8f50634_1 | ||
| x264=1!152.20180806=h14c3975_0 | ||
| xorg-kbproto=1.0.7=h14c3975_1002 | ||
| xorg-libice=1.0.10=h516909a_0 | ||
| xorg-libsm=1.2.3=h84519dc_1000 | ||
| xorg-libx11=1.6.8=h516909a_0 | ||
| xorg-libxau=1.0.9=h14c3975_0 | ||
| xorg-libxdmcp=1.1.3=h516909a_0 | ||
| xorg-libxext=1.3.4=h516909a_0 | ||
| xorg-libxrender=0.9.10=h516909a_1002 | ||
| xorg-renderproto=0.11.1=h14c3975_1002 | ||
| xorg-xextproto=7.3.0=h14c3975_1002 | ||
| xorg-xproto=7.0.31=h14c3975_1007 | ||
| xz=5.2.4=h14c3975_4 | ||
| zeromq=4.3.2=he1b5a44_2 | ||
| zlib=1.2.11=h7b6447c_3 | ||
| zstd=1.4.0=h3b9ef0a_0 |
Oops, something went wrong.