Label Recognition

Label Recognition provides a MATLAB application that is capable of recognising expiry dates within a natural image and displaying them back to the user.

The application utilises a fully documented class, LabelRecogniser, that implements the expiry date detection and recognition algorithm that is provided with this repository.

Prerequisites

This project requires MATLAB, a proprietary language developed by MathWorks, for execution.

It is recommended to use the latest stable release of MATLAB. The performance of releases preceding MATLAB 2018a is not guaranteed.

Please check with your institution/organisation to see if you are offered a free MATLAB license. Otherwise, see the licensing options available.

Installation

Application

1. Navigate to the latest release of this repository.

2. Download the Label-Recognition.mlappinstall file.

3. Open the file and follow the instructions until installation completes.

Class

1. Extract the zip file or clone the repository into the desired directory.

2. Add the parent directory (Label-Recognition) to the path by:

   • Going to Home → Environment → Set Path, or
   • Running addpath [directory] in the Command Window, or
   • Having the parent directory selected as the Current Folder.

3. Use in your own MATLAB code.

Note: Separating the LabelRecogniser.m file into its own directory may improve path performance.

Usage

Application

Ensure that the current folder does not contain LabelRecognition.m or LabelRecogniser.m files since they could shadow the app entry point.

1. Go to Apps Tab → My Apps → Label Recognition.

2. Use the GUI window to perform actions such as:

   • Selecting an image
   • Performing date recognition
   • Configuring detection parameters
   • Viewing recognised dates
   • Outputting dates to a file

Auto-Reflow is used to ensure that the app optimally scales to the desired screen size.

Class

1. Initialise a LabelRecogniser object with an image. This can be either a file path or an image matrix.

   lr = LabelRecogniser('samples/good1.jpeg');

2. Call recogniseDates() on the LabelRecogniser object to identify expiry dates

   dates = lr.recogniseDates()
   
   dates = 
       "25 MAR"

   Alternatively, candidate bounding boxes may be returned too.

   [dates, bboxes] = lr.recogniseDates()
   
   dates = 
       "25 MAR"
   
   bboxes = 
       252.8333    1.0000  106.3333   48.0000
       374.1667  241.5000  165.0000   56.0000
       ...       ...       ...        ...

Further help is provided by using the Command Window to execute:

doc LabelRecogniser

Demonstration

Fig 1: MATLAB app successfully identifying the expiry date from an image.

Fig 2: MATLAB app automatically reflowing content to fit onto a smaller screen.

Fig 3: MATLAB app displaying Clear error messages that suggest possible fixes.

Alternatively, a demonstration of the LabelRecogniser class functionality is provided by the main.m script. Execute it in the Command Window using:

main

Standalone Executable

A standalone executable version of this project was planned to allow non-MATLAB users to make use of the system. However, due to unforeseen complications relating to C/C++ code generation, this task will take longer than expected.

This is due to several functions such as wiener2, detectMSERFeatures, imresize, imcrop and ocr having quirks that will require a major rework of the LabelRecogniser class due to incompatibilities with the MATLAB Compiler.

Additional details for this problem can be found on the issues page. Any contribution will be greatly appreciated.

Author

Daniel Turner - turnerdaniel

References

Stroke Width Transform Algorithm:

Li, Y. and Lu, H. (2012) Scene Text Detection via Stroke Width. In: 21st International Conference on Pattern Recognition (ICPR 2012), Tsukuba, Japan, 11-15 November. IEEE, 681–684. Available from https://ieeexplore.ieee.org/document/6460226 [accessed February 2019].

Candidate Text Grouping Algorithm:

MathWorks (2019) Automatically Detect and Recognize Text in Natural Images. Available from https://uk.mathworks.com/help/vision/examples/automatically-detect-and-recognize-text-in-natural-images.html [accessed February 2019].