Ecosyste.ms: Awesome

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

Awesome Lists | Featured Topics | Projects

https://github.com/autonise/CRAFT-Remade

Implementation of CRAFT Text Detection
https://github.com/autonise/CRAFT-Remade

craft detection ocr pytorch pytorch-implementation text-detection weak-supervision

Last synced: 2 months ago
JSON representation

Implementation of CRAFT Text Detection

Host: GitHub
URL: https://github.com/autonise/CRAFT-Remade
Owner: autonise
License: mit
Created: 2019-07-23T11:08:39.000Z (over 5 years ago)
Default Branch: master
Last Pushed: 2023-07-06T21:42:20.000Z (over 1 year ago)
Last Synced: 2024-08-02T11:15:27.006Z (5 months ago)
Topics: craft, detection, ocr, pytorch, pytorch-implementation, text-detection, weak-supervision
Language: Python
Homepage:
Size: 2.66 MB
Stars: 191
Watchers: 21
Forks: 46
Open Issues: 8
Metadata Files:
- Readme: ReadMe.md
- License: LICENSE

Awesome Lists containing this project

README

# Re-Implementing CRAFT-Character Region Awareness for Text Detection

## Objective

- [X] Reproduce weak-supervision training as mentioned in the paper https://arxiv.org/pdf/1904.01941.pdf
- [ ] Generate character bbox on all the popular data sets.
- [ ] Expose pre-trained models with command line interface to synthesize results on custom images

## Clone the repository

git clone https://github.com/autonise/CRAFT-Remade.git
cd CRAFT-Remade

## Option 1: Conda Environment Installation
conda env create -f environment.yml
conda activate craft

## Option 2: Pip Installation
pip install -r requirements.txt

## Running on custom images

Put the images inside a folder.

Get a pre-trained model from the pre-trained model list (Currently only strong supervision using SYNTH-Text available)

Run the command -

python main.py synthesize --model=./model/final_model.pkl --folder=./input

## Results

| Dataset | Recall | Precision | F-score |
|---------|--------|-----------|---------|
| ICDAR2013 | TBD | TBD | 0.8201(Improving) |
| ICDAR2015 | TBD | TBD | TBD(Coming Soon) |
| ICDAR2017 | TBD | TBD | TBD(Coming Soon) |
| Total Text | TBD | TBD | TBD(Coming Soon) |
| MS COCO | TBD | TBD | TBD(Coming Soon) |

## Pre-trained models

### Strong Supervision

SynthText(CRAFT Model) - https://drive.google.com/open?id=1QH0B-iQ1Ob2HkWCQ2bVCsLPwVSmbcSgN

SynthText(ResNet-UNet Model) - https://drive.google.com/file/d/1qnLM_iMnR1P_6OLoUoFtrReHe4bpFW3T

Original Model by authors - https://drive.google.com/open?id=1ZQE0tK9498RhLcXwYRgod4upmrYWdgl9

### Weak Supervision

- [ ] ICDAR 2013 - https://drive.google.com/file/d/16dl08NtuXX0_8N-vi6uNqQHe83eIvy3U/view?usp=sharing
- [ ] ICDAR 2015 - In Progress
- [ ] ICDAR 2017 - In Progress
- [ ] Total Text - yet_to_be_completed
- [ ] MS-COCO - yet_to_be_completed

## Pre-generated on popular data sets

- [ ] ICDAR 2013 - In Progress
- [ ] ICDAR 2015 - In Progress
- [ ] ICDAR 2017 - yet_to_be_completed
- [ ] Total Text - yet_to_be_completed
- [ ] MS-COCO - yet_to_be_completed

## How to train the model from scratch

### Strong Supervision on Synthetic dataset

Download the pre-trained model on Synthetic dataset at https://drive.google.com/open?id=1qnLM_iMnR1P_6OLoUoFtrReHe4bpFW3T

Otherwise if you want to train from scratch

Run the command -

python main.py train_synth

To test your model on SynthText, Run the command -

python main.py test_synth --model /path/to/model

### Weak Supervision

#### First Pre-Process your dataset

*Currently Supported - [IC13, IC15]

The assumed structure of the dataset is

.
├── Generated (This folder will contain the weak-supervision intermediate targets)
└── Images
├── test
│   ├── img_1.jpg
│   ├── img_2.jpg
│   ├── img_3.jpg
│   ├── img_4.jpg
│   └── img_5.jpg
│   └── ...
├── test_gt.json (This can be generated using the pre_process function described below)
├── train
│   ├── img_1.jpg
│   ├── img_2.jpg
│   ├── img_3.jpg
│   ├── img_4.jpg
│   └── img_5.jpg
│   └── ...
└── train_gt.json (This can be generated using the pre_process function described below)

To generate the json files for IC13 -

In config.py change the corresponding values

'ic13': {
'train': {
'target_json_path': None, --> path to where you want the target json file (Images/train_gt.json)
'target_folder_path': None, --> path to where you downloaded the train gt (ch2_training_localization_transcription_gt)
},
'test': {
'target_json_path': None, --> path to where you want the target json file (Images/test_gt.json)
'target_folder_path': None, --> path to where you downloaded the train gt (Challenge2_Test_Task1_GT)
}

Run the command -

python main.py pre_process --dataset IC13

To generate the json files for IC15 -

In config.py change the corresponding values

'ic15': {
'train': {
'target_json_path': None, --> path to where you want the target json file (Images/train_gt.json)
'target_folder_path': None, --> path to where you downloaded the train gt (ch4_training_localization_transcription_gt)
},
'test': {
'target_json_path': None, --> path to where you want the target json file (Images/test_gt.json)
'target_folder_path': None, --> path to where you downloaded the train gt (Challenge4_Test_Task1_GT)
}

Run the command -
python main.py pre_process --dataset IC15

#### Second Train your model based on weak-supervision

Run the command -

python main.py weak_supervision --model /path/to/strong/supervision/model --iterations

This will train the weak supervision model for the number of iterations you specified