Tensorflow-based CNN+LSTM trained with CTC-loss for OCR

Overview

Overview

This collection demonstrates how to construct and train a deep, bidirectional stacked LSTM using CNN features as input with CTC loss to perform robust word recognition.

The model is a straightforward adaptation of Shi et al.'s CRNN architecture (arXiv:1507.0571). The provided code downloads and trains using Jaderberg et al.'s synthetic data (IJCV 2016), MJSynth.

Notably, the model achieves a lower test word error rate (1.82%) than CRNN when trained and tested on case-insensitive, closed vocabulary MJSynth data.

Written for Python 2.7. Requires TensorFlow >=1.10 (deprecation warnings exist for TF>1.10, but the code still works).

The model and subsequent experiments are more fully described in Weinman et al. (ICDAR 2019)

Structure

The model as built is a hybrid of Shi et al.'s CRNN architecture (arXiv:1507.0571) and the VGG deep convnet, which reduces the number of parameters by stacking pairs of small 3x3 kernels. In addition, the pooling is also limited in the horizontal direction to preserve resolution for character recognition. There must be at least one horizontal element per character.

Assuming one starts with a 32x32 image, the dimensions at each level of filtering are as follows:

Layer Op KrnSz Stride(v,h) OutDim H W PadOpt
1 Conv 3 1 64 30 30 valid
2 Conv 3 1 64 30 30 same
Pool 2 2 64 15 15
3 Conv 3 1 128 15 15 same
4 Conv 3 1 128 15 15 same
Pool 2 2,1 128 7 14
5 Conv 3 1 256 7 14 same
6 Conv 3 1 256 7 14 same
Pool 2 2,1 256 3 13
7 Conv 3 1 512 3 13 same
8 Conv 3 1 512 3 13 same
Pool 3 3,1 512 1 13
9 LSTM 512
10 LSTM 512

To accelerate training, a batch normalization layer is included before each pooling layer and ReLU non-linearities are used throughout. Other model details should be easily identifiable in the code.

The default training mechanism uses the ADAM optimizer with learning rate decay.

Differences from CRNN

Deeper early convolutions

The original CRNN uses a single 3x3 convolution in the first two conv/pool stages, while this network uses a paired sequence of 3x3 kernels. This change increases the theoretical receptive field of early stages of the network.

As a tradeoff, we omit the computationally expensive 2x2x512 final convolutional layer of CRNN. In its place, this network vertically max pools over the remaining three rows of features to collapse to a single 512-dimensional feature vector at each horizontal location.

The combination of these changes preserves the theoretical receptive field size of the final CNN layer, but reduces the number of convolution parameters to be learned by 15%.

Padding

Another important difference is the lack of zero-padding in the first convolutional layer, which can cause spurious strong filter responses around the border. By trimming the first convolution to valid regions, this model erodes the outermost pixel of values from the response filter maps (reducing height from 32 to 30 and reducing the width by two pixels).

This approach seems preferable to requiring the network to learn to ignore strong Conv1 responses near the image edge (presumably by weakening the power of filters in subsequent convolutional layers).

Batch normalization

We include batch normalization after each pair of convolutions (i.e., after layers 2, 4, 6, and 8 as numbered above). The CRNN does not include batch normalization after its first two convolutional stages. Our model therefore requires greater computation with an eye toward decreasing the number of training iterations required to reach converegence.

Subsampling/stride

The first two pooling stages of CRNN downsample the feature maps with a stride of two in both spatial dimensions. This model instead preserves sequence length by downsampling horizontally only after the first pooling stage.

Because the output feature map must have at least one timeslice per character predicted, overzealous downsampling can make it impossible to represent/predict sequences of very compact or narrow characters. Reducing the horizontal downsampling allows this model to recognize words in narrow fonts.

This increase in horizontal resolution does mean the LSTMs must capture more information. Hence this model uses 512 hidden units, rather than the 256 used by the CRNN. We found this larger number to be necessary for good performance.

Training

To completely train the model, you will need to download the mjsynth dataset and pack it into sharded TensorFlow records. Then you can start the training process, a tensorboard monitor, and an ongoing evaluation thread. The individual commands are packaged in the accompanying Makefile.

make mjsynth-download
make mjsynth-tfrecord
make train &
make monitor &
make test

To monitor training, point your web browser to the url (e.g., (http://127.0.1.1:8008)) given by the Tensorboard output.

Note that it may take 4-12 hours to download the complete mjsynth data set. A very small set (0.1%) of packaged example data is included; to run the small demo, skip the first two lines involving mjsynth.

With a GeForce GTX 1080, the demo takes about 20 minutes for the validation character error to reach 45% (using the default parameters); at one hour (roughly 7000 iterations), the validation error is just over 20%.

With the full training data, by one million iterations the model typically converges to around 5% training character error and 27.5% word error.

Checkpoints

Pre-trained model checkpoints at DOI:11084/23328 are used to produce results in the following paper:

Weinman, J. et al. (2019) Deep Neural Networks for Text Detection and Recognition in Historical Maps. In Proc. ICDAR.

Testing

The evaluate script (src/evaluate.py) streams statistics for one batch of validation (or evaluation) data. It prints the iteration, evaluation batch loss, label error (percentage of characters predicted incorrectly), and the sequence error (percentage of words—entire sequences—predicted incorrectly).

The test script (src/test.py) tallies statistics, finally normalizing for all data. It prints the loss, label error, total number of labels, sequence error, total number of sequences, and the label error rate and sequence error rate.

Validation

To see the output of a small set of instances, the validation script (src/validation.py) allows you to load a model and read an image one at a time via the process's standard input and print the decoded output for each. For example

cd src ; python validate.py < ~/paths_to_images.txt

Alternatively, you can run the program interactively by typing image paths in the terminal (one per line, type Control-D when you want the model to run the input entered so far).

Configuration

There are many command-line options to configure training parameters. Run train.py or test.py with the --help flag to see them or inspect the scripts. Model parameters are not command-line configurable and need to be edited in the code (see src/model.py).

Dynamic training data

Dynamic data can be used for training or testing by setting the --nostatic_data flag.

You can use the --ipc_synth boolean flag [default=True] to determine whether to use single-threaded or a buffered, multiprocess synthesis.

The --synth_config_file flag must be given with --nostatic_data.

The MapTextSynthesizer library supports training with dynamically synthesized data. The relevant code can be found within MapTextSynthesizer/tensorflow/generator

Using a lexicon

By default, recognition occurs in "open vocabulary" mode. That is, the system observes no constraints on producing the resulting output strings. However, it also has a "closed vocabulary" mode that can efficiently limit output to a given word list as well as a "mixed vocabulary" mode that can produce either a vocabulary word from a given word list (lexicon) or a non-vocabulary word, depending on the value of a prior bias for lexicon words.

Using the closed or mixed vocabulary modes requires additional software. This repository is connected with a fork of Harald Scheidl's CTCWordBeamSearch, obtainable as follows:

git clone https://github.com/weinman/CTCWordBeamSearch
cd CTCWordBeamSearch
git checkout var_seq_len

Then follow the build instructions, which may be as simple as running

cd cpp/proj
./buildTF.sh

To use, make sure CTCWordBeamSearch/cpp/proj (the directory containing TFWordBeamSearch.so) is in the LD_LIBRARY_PATH when running test.py or validate.py (in this repository).

API Notes

This version uses the TensorFlow (v1.14) Dataset for fast I/O. Training, testing, validation, and prediction use a custom Estimator.

Citing this work

Please cite the following paper if you use this code in your own research work:

@inproceedings{ weinman19deep,
    author = {Jerod Weinman and Ziwen Chen and Ben Gafford and Nathan Gifford and Abyaya Lamsal and Liam Niehus-Staab},
    title = {Deep Neural Networks for Text Detection and Recognition in Historical Maps},
    booktitle = {Proc. IAPR International Conference on Document Analysis and Recognition},
    month = {Sep.},
    year = {2019},
    location = {Sydney, Australia},
    doi = {10.1109/ICDAR.2019.00149}
} 

Acknowledgment

This work was supported in part by the National Science Foundation under grant Grant Number 1526350.

Owner
Jerod Weinman
Associate Professor of Computer Science
Jerod Weinman
An interactive interface for using OpenCV's GrabCut algorithm for image segmentation.

Interactive GrabCut An interactive interface for using OpenCV's GrabCut algorithm for image segmentation. Setup Install dependencies: pip install nump

Jason Y. Zhang 16 Oct 10, 2022
End-to-end pipeline for real-time scene text detection and recognition.

Real-time-Scene-Text-Detection-and-Recognition-System End-to-end pipeline for real-time scene text detection and recognition. The detection model use

Fangneng Zhan 89 Aug 04, 2022
An expandable and scalable OCR pipeline

Overview Nidaba is the central controller for the entire OGL OCR pipeline. It oversees and automates the process of converting raw images into citable

81 Jan 04, 2023
nofacedb/faceprocessor is a face recognition engine for NoFaceDB program complex.

faceprocessor nofacedb/faceprocessor is a face recognition engine for NoFaceDB program complex. Tech faceprocessor uses a number of open source projec

NoFaceDB 3 Sep 06, 2021
Hand gesture detection project with aweome UI implementation.

an awesome hand gesture detection project for you to be creative! Imagination is the limit to do with this project.

AR Ashraf 39 Sep 26, 2022
Layout Analysis Evaluator for the ICDAR 2017 competition on Layout Analysis for Challenging Medieval Manuscripts

LayoutAnalysisEvaluator Layout Analysis Evaluator for: ICDAR 2019 Historical Document Reading Challenge on Large Structured Chinese Family Records ICD

17 Dec 08, 2022
Text page dewarping using a "cubic sheet" model

page_dewarp Page dewarping and thresholding using a "cubic sheet" model - see full writeup at https://mzucker.github.io/2016/08/15/page-dewarping.html

Matt Zucker 1.2k Dec 29, 2022
Memory tests solver with using OpenCV

Human Benchmark project This project is OpenCV based programs which are puzzle solvers for 7 different games for https://humanbenchmark.com/. made as

Bahadır Araz 24 Dec 27, 2022
A buffered and threaded wrapper for the OpenCV VideoCapture object. Can speed up video decoding significantly. Supports

A buffered and threaded wrapper for the OpenCV VideoCapture object. Can speed up video decoding significantly. Supports "with"-syntax.

Patrice Matz 0 Oct 30, 2021
A curated list of promising OCR resources

Call for contributor(paper summary,dataset generation,algorithm implementation and any other useful resources) awesome-ocr A curated list of promising

wanghaisheng 1.6k Jan 04, 2023
APS 6º Semestre - UNIP (2021)

UNIP - Universidade Paulista Ciência da Computação (CC) DESENVOLVIMENTO DE UM SISTEMA COMPUTACIONAL PARA ANÁLISE E CLASSIFICAÇÃO DE FORMAS Link do git

Eduardo Talarico 5 Mar 09, 2022
Character Segmentation using TensorFlow

Character Segmentation Segment characters and spaces in one text line,from this paper Chinese English mixed Character Segmentation as Semantic Segment

26 Aug 25, 2022
Code for the head detector (HeadHunter) proposed in our CVPR 2021 paper Tracking Pedestrian Heads in Dense Crowd.

Head Detector Code for the head detector (HeadHunter) proposed in our CVPR 2021 paper Tracking Pedestrian Heads in Dense Crowd. The head_detection mod

Ramana Subramanyam 76 Dec 06, 2022
Learning Camera Localization via Dense Scene Matching, CVPR2021

This repository contains code of our CVPR 2021 paper - "Learning Camera Localization via Dense Scene Matching" by Shitao Tang, Chengzhou Tang, Rui Hua

tangshitao 65 Dec 01, 2022
Handwritten Number Recognition using CNN and Character Segmentation

Handwritten-Number-Recognition-With-Image-Segmentation Info About this repository This Repository is aimed at reading handwritten images of numbers an

Sparsha Saha 17 Aug 25, 2022
Fatigue Driving Detection Based on Dlib

Fatigue Driving Detection Based on Dlib

5 Dec 14, 2022
A simple demo program for using OpenCV on Android

Kivy OpenCV Demo A simple demo program for using OpenCV on Android Build with: buildozer android debug deploy run Run (on desktop) with: python main.p

Andrea Ranieri 13 Dec 29, 2022
Fine tuning keras-ocr python package with custom synthetic dataset from scratch

OCR-Pipeline-with-Keras The keras-ocr package generally consists of two parts: a Detector and a Recognizer: Detector is responsible for creating bound

Eugene 1 Jan 05, 2022
RRD: Rotation-Sensitive Regression for Oriented Scene Text Detection

RRD: Rotation-Sensitive Regression for Oriented Scene Text Detection For more details, please refer to our paper. Citing Please cite the related works

Minghui Liao 102 Jun 29, 2022
Fun program to overlay a mask to yourself using a webcam

Superhero Mask Overlay Description Simple project made for fun. It consists of placing a mask (a PNG image with transparent background) on your face.

KB Kwan 10 Dec 01, 2022