Repository files navigation

This is a repository forked fromargman/EAST for theICPR MTWI 2018 CHALLENGE.
Origin Repository:argman/EAST - A tensorflow implementation of EAST text detector
Origin Author:argman

Author:Haozheng Li
Email:sai-2008@qq.com orakaHaozhengLi@gmail.com

1. Transform
2. Models
3. Demo
4. Train
5. Test
6. Results

Some data in the dataset is abnormal, just likeICPR MTWI 2018[BaiduYun]. Abnormal means that the ground true labels are anticlockwise, or the images are not in 3 channels. Then errors like'poly in wrong direction' will occur while usingargman/EAST.

So I wrote a matlab program to check and transform the dataset. The program named <transform.m> is in the folder 'data_transform/' and its parameters are descripted as bellow:

icpr_img_folder = 'image_9000\';                   %origin imagesicpr_txt_folder = 'txt_9000\';                     %origin ground true labelsicdar_img_folder = 'ICPR2018\';                    %transformed imagesicdar_gt_folder = 'ICPR2018\';                     %transformed ground true labelsicdar_img_abnormal_folder = 'ICPR2018_abnormal\';  %images not in 3 channels, which give errors in argman/EASTicdar_gt_abnormal_folder = 'ICPR2018_abnormal\';   %transformed ground true labels%images and ground true labels files must be renamed as <img_1>, <img_2>, ..., <img_xxx> while using argman/EASTfirst_index =  0;                                  %first index of the datasettransform_list_name = 'transform_list.txt';        %file name of the rename list

Note: For abnormal images not in 3 channels, please transform them to normal ones through other tools likeFormat Factory. Then add the right data to the <icdar_img_folder> and <icdar_gt_folder>, so finally you get a whole normal dataset which has been checked and transformed.

1. Resnet_V1_50 Models trained onICPR MTWI 2018 (train): [100k steps] [500k steps] [1035k steps]
2. Resnet_V1_101 Models trained onICPR MTWI 2018 (train) +ICDAR 2017 MLT (train + val) +RCTW-17 (train): [100k steps]
3. Resnet_V1_101 Models pre-trained on Models-2, then trained on justICPR MTWI 2018 (train): [987k steps]
4. (Inargman/EAST) Resnet_V1_50 Models trained on ICDAR 2013 (train) + ICDAR 2015 (train): [50k steps]
5. (Inargman/EAST) Resnet_V1_50 Models provided by tensorflow slim: [slim_resnet_v1_50]

Download the pre-trained models and run:

python run_demo_server.py --checkpoint-path models/east_icpr2018_resnet_v1_50_rbox_1035k/

Then Openhttp://localhost:8769 for the web demo server, or get the results in 'static/results/'.
Note: Seeargman/EAST#demo for more details.

Prepare the training set and run:

python multigpu_train.py --gpu_list=0 --input_size=512 --batch_size_per_gpu=8 \--checkpoint_path=models/east_icpr2018_resnet_v1_50_rbox/ \--text_scale=512 --training_data_path=data/ICPR2018/ --geometry=RBOX \--learning_rate=0.0001 --num_readers=18 --max_steps=50000

Note 1: Images and ground true labels files must be renamed as <img_1>, <img_2>, ..., <img_xxx> while usingargman/EAST. Please see the examples in the folder 'training_samples/'.
Note 2: If--restore=True, training will restore from checkpoint and ignore the--pretrained_model_path. If--restore=False, training will delete checkpoint and initialize with the--pretrained_model_path (if exists).
Note 3: Seeargman/EAST#train for more details.

Names of the images inICPR MTWI 2018 are abnormal. Like <LB1gXi2JVXXXXXUXFXXXXXXXXXX.jpg> but not <img_10001.jpg>. Then errors will occur while usingargman/EAST#test.

So I wrote two matlab programs to rename and inversely rename the dataset. Before evaluating, run the program named <rename.m> to make names of the images normal. This program is in the folder 'data_transform/' and its parameters are descripted as bellow:

icpr_img_folder = 'image_10000\';                      %origin imagesicdar_img_folder = 'ICPR2018_test\';                   %transformed imagesicdar_img_abnormal_folder = 'ICPR2018_test_abnormal\'; %images not in 3 channels, which give errors in argman/EASTicpr_count =  10000;                                   %first index of the datasetrename_list_name = 'rename_list.txt';                  %file name of the rename list

Note: Just like <transform.m>, please transform abnormal images through other tools likeFormat Factory.

After you have prepared the test set, run:

python eval.py --test_data_path=data/ICPR2018/ --gpu_list=0 \--checkpoint_path=models/east_icpr2018_resnet_v1_50_rbox_1035k/ --output_dir=results/1035k/

Then get the results in 'results/'.

Finally inversely rename the result labels files from <img_10001.txt> to <LB1gXi2JVXXXXXUXFXXXXXXXXXX.txt> according to the rename list generated by <rename.m>. Run the program named <rename_inverse.m> which is in the folder 'data_transform/' and its parameters are descripted as bellow:

rename_list_name = 'rename_list.txt';  %file name of the rename listicpr_img_folder = 'image_10000\';      %origin imagesicpr_txt_folder = 'results\';          %result labels files generated by argman/EASTicdar_gt_folder = 'txt_10000\';        %inversely renamed result labels files

Then zip the results in 'txt_10000/' and submit it to theICPR MTWI 2018 CHALLENGE.

Finally our model <east_icpr2018_resnet_v1_50_rbox_1035k> rank 31 in theICPR MTWI 2018 CHALLENGE:

Here are some results onICPR MTWI 2018: