BIQE OCR Server
BIQE OCR Server complements modern high-speed document scanners with a highly optimized subsystem that applies optical character recognition (OCR) and Mixed Raster Content (MRC) compression to many different file types.
With BIQE OCR Server you can easily create highly compressed and searchable PDFs without manual intervention. It has been build around a sophisticated Hot Folder system that processes files as soon as they hit the source directory. As such, BIQE OCR Server can easily be integrated into existing scanning workflows.
The exceptional speed offered by BIQE OCR Server is obtained by full parallelization of processes, i.e. multiple pages are processed concurrently. Consequently, performance depends heavily on the machine that runs the software.
BIQE OCR Server is offered in the following configurations:
- 4 cores / 8 threads (OEM only)
- 8 cores / 16 threads
- 16 cores / 32 threads
- 28 cores / 56 threads
- unlimited cores / unlimited threads
1.1 Which license do I need?
Determining which license you need depends on two factors:
- The computer system you want to use to run BIQE OCR Server.
- The performance required
BIQE OCR Server is physically limited by the number of cores and threads (usually 2 threads for each core) available in the computer system. Hence, it does not make sense to buy a more expensive license (allowing more threads to be used) if the computer system does not contain these additional threads.
Usually, the number of threads is double the number of available cores (please check your processors documentation). To inspect how many threads your current system has, open Device Manager in Windows and inspect Processors in the overview. It lists a processor for each available thread.
Another important consideration in determining which license (and perhaps which hardware) you need is desired performance. By using a trial version of BIQE OCR Server, it is easy to figure out how much time your system requires to process a representative selection of typical documents in your workflow.
If it takes 10 seconds on average to process a page and your system is capable of running 16 threads in parallel, it processes about 96 pages per minute (60 seconds / 10 seconds * 16 threads). If you are using a document scanner that scans 200 pages per minute, you need to upgrade both your hardware and your license to that level to keep up with your scanner’s speed. Otherwise you have to wait after your document scanner has finished.
To make the calculation, divide the number of pages per minute by the number of pages one thread can process per minute. The result of that calculation yields the number of threads required to process the pages produced by the document scanner without delay.
E.g. 200 pages per minute / 10 pages per thread per minute = 20 threads.
Consequently, you will need a system that has at least 10 cores with two threads each. The only way to know how fast one thread can process one page is by trying it out with a trial version of BIQE OCR Server. However, modern systems (0-3 years old) usually process pages between 5-15 seconds per thread if both OCR and MRC compression are enabled.
Open the installation file and follow the procedure to install BIQE OCR Server. You need administrator rights to complete the installation procedure successfully.
After reading and accepting the Terms of Service and the End User License Agreement, you press Install.
Windows will ask to do the installation with administrator rights. After confirmation, the installation procedure starts and BIQE OCR Server will be installed including some required packages.
Upon finishing the installation successfully, you will be asked to restart your computer.
1.3 License activation
When you start BIQE OCR Server for the first time, it might indicate that it is not activated yet. Without activation, it is not possible to use the software. However, the activation procedure is smooth and convenient. To activate your product, click About in the menu.
This opens a window that indicates that no license is activated.
To activate your license, enter your personal data and the license number you received.
Click activate and a message appears: Your license has been successfully activated
You will find your license details at the About page. You are ready to use BIQE OCR Server.
1.4 Check activation status
To check the activation status, click About in the menu bar.
If activated correctly, the about window will show the license status, its type, and the maximum number of threads allowed under the current license.
Press OK to close the window.
2 Get started
This chapter provides a condensed overview of BIQE OCR Server, without going into detail. Detailed descriptions are to be found in subsequent chapters. Reading this chapter helps to understand the main components and the general workflow.
2.1 The user interface
- Directory management
- Statistic area
2.2 Directory management
In the first section (A), three directories have to be defined:
- Source directory
- Destination directory
- Backup directory
The Source directory contains the files to be processed by BIQE OCR Server and is the location of the Hot Folder. The Destination directory is the place where files will be written to after processing. The Backup directory collects the original input files in case the user activates the Backup option under Settings.
All three directories should be defined before starting BIQE OCR Server.
The second section (B) provides instant access to what is going on while BIQE OCR Server is running. The first progress bar counts the number of pages processed, while the second bar counts the number of files processed. In case every file contains one page only, both bars should be equally progressing. In case of multipage input files, they will differ.
The settings section (C) is the key panel with all definitions of what BIQE OCR Server will actually do. Here you define the actions to be performed as well as the rules to be applied. To get started, you are advised to start with the default settings.
The two main functions of BIQE OCR Server are defined here:
- Applying MRC Compression
- Applying one or two OCR languages
All details about the settings are elaborated in chapter 4.
2.5 Statistic area
The Statistic Area section provides insights into the pages processed. Ordering on one of the different columns allows you to identify outliers easily. E.g. pages with a different width or height, or the page processing time.
2.6 First time set up
To start your first run, you have to fill in the directories first (cf. 2.3). Enter the directory paths by typing them or by pasting them into the path sections.
You can also click on the three bullets and select the folder you want to use.
After assigning the directories, proceed to select your preferences in the Settings window below and click the START button to begin the process.
3 Directory Management
In this chapter, we explain the operation of directories in BIQE OCR Server (A). BIQE provides advanced interactions with existing systems and pipelines. However, it is important to understand how the directories selected interact with given settings.
3.1 Source directory
The Source directory is a single directory in your computer system. This directory functions as a Hot Folder.
Tip: Create the source folder on the fastest available disk, as the fastest reads will increase the performance of the BIQE OCR Server.
The Source directory can be defined in two ways:
- By typing or pasting the path
- By clicking the three bullets next to the input field
To view the contents of the Source directory, click on the URL or use the shortcut: CTRL + F3. This opens the directory in Windows explorer.
Two settings have an effect on the way how BIQE OCR Server deals with the Source directory:
- Remove input folder after processing
- Processing mode
3.1.1 Remove input folder after processing
Delete input folder after processing determines whether files in the Source directory are deleted after processing. If BIQE OCR Server is part of an automatic pipeline, it is advisable to delete the files after processing to prevent the Source directory to become cluttered.
Enabling Delete input folder after processing will remove your input files permanently. Check the Backup option to ensure that your original files are saved!
3.1.2 Processing mode
BIQE OCR Server has three different Hot Folder functionalities regarding the processing of files in the Source Folder. You can select one Hot Folder functionality per process.
- Retain folder structure
- Waiting for signal file
- Instant file processing
Retain folder structure creates an exact copy of the Source directory in terms of structure of (sub)folders and files. It allows you to organize your data before processing them.
Wait for signal file keeps the Hot Folder inactive even after you click the START button until an empty complete.txt file is created or copied into the Source directory.
Instant file processing turns your Source directory into a Hot Folder that processes files as soon as they enter this directory. This can be very useful in automatic pipelines.
Each of the Processing settings are further elaborated in chapter 5.
3.2 Destination directory
The Destination directory is the folder to which all output of BIQE OCR Server is written. Using a fast hard drive for this directory will improve the overall performance of BIQE OCR Server.
3.3 Backup directory
The Backup directory is the folder to which all original input files are written. It is generally advisable to use the Backup option to keep the original files after processing.
This section explains the progress indicators available in BIQE OCR Server (B). In addition to the general progress bars in the main window, BIQE OCR Server provides two other windows to monitor the process: HF’s activity and HF’s control. Both of these additional windows are explained in this chapter.
4.1 Main window progress bars
The main window shows two different progress bars. As soon as a process is started, the system counts the number of files and the pages available in them. The first progress bar keeps track of the number of pages to be processed. This bar provides the most accurate indication about the progress of the process.
The second progress bar counts the number of files to be processed. If each file contains only one page, the two bars synchronize.
Next to these two indicators, two other indicators are available: a directory count of waiting directories in the upper left corner of the Progress section, and an indicator of elapsed time in the down right corner.
Clicking the waiting directories URL opens a new window with the directories listed. They will be processed after the current process ends.
The case of waiting directories happens when files are added to a directory after it has been frozen (cf. 4.2).
4.2 HFs activity
HFs activity provides insight into the Hot Folder’s (HFs) activity. This is relevant in case of an active document instream from e.g. a document scanner. When large multi-page PDFs are produced, it takes time to write the complete file. To avoid the situation where BIQE OCR Server tries to process a file or a folder while they are still incomplete, this Hot Folder system works with a dual structure: folders that undergo change (i.e. files are still being written) and folders that are frozen because the creation of (sub-)folders and files has been completed.
The parameter to control the time span that no more (sub)folders and files are added and the process can be completed can be set with the Time span freeze folder. The integer entered refers to the number of seconds that must elapse without changing the folder before the folder is processed (frozen). Any change after the set time span is treated as a new process.
With HFs activity, you get a real-time insight into which directories are frozen and which ones not.
4.3 HFs control
HFs control screen gives you insight into the directories that are currently being processed and those that are finished. This screen is especially useful if BIQE OCR Server is running in an automatic pipeline. It provides insight into all processes that have been started since the START button has been pressed.
In the settings section (C), you define the specific requirements for the task you want to perform. In this section, they will be explained one by one.
5.1 Number of threads
BIQE OCR Server was created with a focus on maximum performance. In the processing of files, all available threads of your system can be used. The number of threads available depends on two factors: the number of physical cores (usually 2 threads each) available in the computer system and the license.
The maximum number of threads available is determined by the license, but the actual number is determined by your computer system. If you bought a license for 24 threads, but your system has 8 cores with 16 threads, then the maximum number of threads to be used is 16 threads.
If no other (demanding) processes besides BIQE OCR Server are running, then you can safely use the maximum number of threads available. If BIQE OCR Server is used in an active production pipeline where the same computer system is actively receiving files from e.g. a document scanner, it is recommended to stay at least 2 threads below the maximum, to leave capacity to the other process.
5.2 MRC compression
One of the key functionalities of BIQE OCR Server is its capability to apply Mixed Raster Content (MRC) Compression to input files. BIQE OCR Server offer this functionality without limitation.
Until PDF 1.4, JPEG compression has been used, while in later versions Jp2 compression has been applied. Although the common JPEG/Jp2 compression works well with color images, it performs rather poorly on text, resulting in unpleasant artifacts. Ideally, text should have sharp edges to be optimally readable.
MRC Compression has two major advantages over JPEG/Jp2 compression:
- it significantly reduces the size of the PDF file
- it keeps the text readable and clean
As a result, you need less storage capacity, and transferring files requires much less time because the files you send are smaller.
How does MRC work? MRC Compression is divided into three layers: A binary layer (text), which is stored in high resolution using the JBIG2 algorithm; A highly compressed background layer, which contains the background image after splitting the contents of the previous layer, and a foreground layer, which contains the colors of the binary layer. The final file size reduction rate depends on the Quality and Resolution you choose.
5.3 Preserve smoothing
Preserve Smoothing is a compression algorithm that smooths the edges of letters, making the letter more readable on the screen. This setting is highly recommended for text documents.
5.4 Create a multi-page PDF for a folder
Enabling this setting instructs BIQE OCR Server to treat a directory with files as one document. As a result, one single multi-page PDF will be created from all files available within the directory.
5.5 PDF types
BIQE OCR Server features all common PDF types. By default, BIQE Server chooses the PDF version 1.7, but you can also select another PDF type and version to meet your requirements.
The Quality option sets the compression percentage relative to the original. 1 indicates the highest compression rate with the lowest image quality; 100 is uncompressed and preserves the highest image quality. For text documents, we recommend a compression factor of 15 to have a good balance between compression and image quality. Documents with images, will require a higher quality to keep a convincing result.
Use source keeps the original dpi with the exception of documents with a resolution higher than 600 dpi. They are down sampled to 300 dpi. This value can be upscaled or downscaled according to your needs. Currently the following common values can be used:
- 96 dpi
- 120 dpi
- 150 dpi
- 300 dpi
5.8 Exclude by size
BIQE OCR Server allows you to exclude images with certain dimensions from OCR. You can both exclude images that are below a certain threshold (<) or images that are above a certain threshold (>). Each rule can be set individually by checking the checkboxes.
BIQE OCR Server allows you to use up to two OCR languages simultaneously. For a searchable PDF, you must select the OCR checkbox and choose one or two OCR languages.
BIQE OCR Server uses Tesseract 4.1. The great advantage of using an Open Source OCR Engine is that you can add new languages and train your own Tesseract language models.
5.9.1 Adding additional language models
Which languages can be added can be found in the list here. Languages mentioned can be downloaded here. Another list that can be consulted is the list with supposedly the best models for Tesseract 4 can be found here. However, they are known to be slower than the regular models.
To add a language to BIQE OCR Server, you will need to find the directory where the languages are installed. This is usually C:\Program Files\BIQE-OCR-server.
Within the main directory, you add the model to both tessdata and tessdataGdPicture. Restart BIQE OCR Server and the newly added languages are listed in the OCR language selection menu.
5.9.2 Training your own language models
Training your own Tesseract models requires a lot of technical skills. Please read the Tesseract documentation here to get a better idea of the procedure involved. A more accessible example on how to train a Tesseract model can be found here.
After creating the model, you can add it according to the same procedure as adding an existing model (see 5.9.1).
If you check the Backup option, the original files from the Source directory are copied to the set Backup directory.
Tip: We recommend that you always check the Backup option in combination with the Delete input directory after processing.
5.11 Remove input folder after processing
This option causes the Hot Folder (Source directory) to be emptied after processing the files.
If you do not have the Backup option checked, the original files will be permanently deleted!
5.12 Hot folder freezing time span
The Hot folder freezing time waits to process files until the set seconds (maximum 90) have passed. This prevents files from being processed while still being written to the source directory.
The default is 5 seconds. However, if very large files are written to the source directory (>100MB), it is important to increase the number of seconds to a safe zone.
5.13 Create ALTO xml
ALTO stands for Analyzed Layout and Text Object. The document consists of both the recognized text and the precise text positions of the text on the image.
This allows you to inspect the text layer easily, without the need to extract it from the PDF.
Be aware, however, that enabling this function will create an XML file in addition to the PDF file, resulting in 2 output files for one input file.
5.14 ALTO xml: Per symbol output
BIQE OCR Server offers, in addition to the ALTO xml option at WORD level, the option to save the ALTO xml at TOKEN level. Enable this checkbox if you have that requirement specifically.
5.15 Processing mode
BIQE OCR Server offers three processing modes that define the behavior of the Hot Folder more precisely.
5.15.1 Retain folder structure
This option preserves the original folder (and subfolder) structure according to the Source folder. After processing, the folder structure of the Destination folder will be identical to that of the Source folder.
If any files are added to the Hot Folder during processing, it is important to set the Hot Folder freeze time properly, especially if the files are large. If the freezing time expires without any more files being added to the Hot Folder, the folder will be locked and all changes will be considered a new process.
5.15.2 Waiting for signal file
In this Hot folder mode, file processing will not begin until an empty complete.txt file is created/copied into the Source directory. This mode does not allow nested folder structures!
This mode is especially useful when input data needs to be processed in batches and when the scanner produces a signal file after scanning a batch of documents.
TIP: For example, a bank can easily process customer data separately because they copy the complete.txt to the Source Directory after processing each customer.
5.15.3 Instant file processing
In this Hot folder mode, file This mode processes every single file as soon as it hits the Hot Folder, as long as the START button is active. This option does not allow for multi-page PDF creation nor does it permit to remove files from the Source directory automatically.after processing each customer.
6 Statistic Area
Finally, BIQE OCR Server provides a real-time interface to understand processed page data (D). In addition to the file name, the page number, total number of pages in the file, width and height, and processing time are listed.
You can click on each column to reorder its content. That way, you can easily detect deviations in time and size. Pages that are processed very quickly usually have little or no text that needs to be recognized. By organizing the processed files by processing time, you can easily find the empty pages.
The full logs are written to an Excel file that can be accessed in the upper right corner or by pressing the shortcut CTRL + L. Log files are written to a separate ’log’ directory in the Destination directory.
BIQE OCR Server
- Unlimited OCR
- Unlimited MRC PDF compression
- Fully scalable according to available cores/threads
- Unique hotfolder processing