A small lightweight HTTP server that converts photos, images and scanned documents to text using optical character recognition by utilizing the power of Google Tesseract.
The easiest way to get started is using pre-built docker images (multi-arch)
$ docker run -p 8884:8884 hertzg/tesseract-server:latestYou can use the service by sending multipart http requests containing
options and file fields.
# Run OCR using english language on file sample.jpg in current directory
$ curl -F "options={\"languages\":[\"eng\"]}" -F file=@sample.jpg http://127.0.0.1:8884/tesseract
{
"data": {
"exit": {
"code": 0,
"signal": null
},
"stderr": "Warning: Invalid resolution 0 dpi. Using 70 instead.\nEstimating resolution as 153\n",
"stdout": " \n\n \n\nThe Life and Work of\nFredson Bowers\n\nby\nG. THOMAS TANSELLE\n\n \n\nN EVERY FIELD OF ENDEAVOR THERE ARE A FEW FIGURES WHOSE AGCOM-\nplishment and influence cause them to be the symbols of their age;\ntheir careers and oeuvres become the touchstones by which the\nfield is measured and its history told. In the related pursuits of\n\nanalytical and descriptive bibliography, textual criticism, and scholarly\nediting, Fredson Bowers was such a figure, dominating the four decades\nafter 1949, when his Principles of Bibliographical Description was pub-\nlished. By 1973 the period was already being called “the age of Bowers”:\nin that year Norman Sanders, writing the chapter on textual scholarship\nfor Stanley Wells's Shakespeare: Select Bibliographies, gave this title to\na section of his essay. For most people, it would be achievement enough\nto ise to such a position in a field as complex as Shakespearean textual\nstudies; but Bowers played an equally important role in other areas.\nEditors of nineteenth-century American authors, for example, would\nalso have to call the recent past “the age of Bowers, as would the writers\nof descriptive bibliographics of authors and presses. His ubiquity in\nthe broad field of bibliographical and textual study, his seemingly com-\nplete possession of it, distinguished him from his illustrious predeces-\nsors and made him the personification of bibliographical scholarship in\nhis time.\n\nWhen in 1969 Bowers was awarded the Gold Medal of the Biblio-\ngraphical Society in London, John Carter’s citation referred to the\nPrinciples as “majestic,” called Bowers’s current projects “formidable,”\nsaid that he had “imposed critical discipline” on the texts of several\nauthors, described Studies in Bibliography as a “great and continuing\nachievement,” and included among his characteristics “uncompromising\nseriousness of purpose” and “professional intensity.” Bowers was not\nunaccustomed to such encomia, but he had also experienced his share of\nattacks: his scholarly positions were not universally popular, and he\nexpressed them with an aggressiveness that almost seemed calculated to\n\n \n\f"
}
}The service provides configurations as cli options. All the options with their
descriptions, types and defaults including some usage examples can be seen using
--help flag.
# Using Docker
$ docker hertzg/tesseract-server:latest --helptesseract-server [options]
A small lightweight http server exposing tesseract as a service.
Options:
--help Show help [boolean]
--version Show version number [boolean]
--pool.default.min Minimum number of processes to keep waiting in each pool [number] [default: 0]
--pool.default.max Maximum number of processes to spawn for each pool after which requests are queued [number] [default: 2]
--pool.default.idleTimeoutMillis Time (in milliseconds) a processes can stay idle in queue before eviction [number] [default: 5000]
--pool.default.evictionRunIntervalMillis Time interval (in milliseconds) between eviction checks [number] [default: 5000]
--http.listen.address Set http listen address [string] [default: "0.0.0.0"]
--http.listen.port Set http listen port [number] [default: 8884]
--http.upload.tmpDir Path to where temp uploads are saved to [string] [default: "/tmp"]
--http.endpoint.status.enable Enable /status endpoint [boolean] [default: true]
--http.endpoint.health.enable Enable /.well-known/health/* endpoints and health checkers [boolean] [default: true]
--http.endpoint.webui.enable Enable Web UI at / [boolean] [default: true]
--http.input.optionsField Multipart field name containing OCR Options [string] [default: "options"]
--http.input.fileField Multipart field name containing OCR file [string] [default: "file"]
--http.output.jsonSpaces Enable json pretty printing and set number of spaces to use for indentation [number] [default: 0]
--processor.lineEndings Set line ending policy [string] [choices: "auto", "lf", "crlf"] [default: "auto"]
Examples:
tesseract-server --http.output.jsonSpaces 2 Enable JSON pretty printing
tesseract-server --http.endpoint.status.enable false Disable Status and Health endpoints
--http.endpoint.health.enable false
References:
GitHub: https://github.com/hertzg/tesseract-server
Discussions: https://github.com/hertzg/tesseract-server/discussions
Issues: https://github.com/hertzg/tesseract-server/issues
Docker images are multi-arch images based on alpine variant of official node
docker images supporting linux/amd64, linux/arm/v6, linux/arm/v7,
linux/arm64/v8, linux/ppc64le and linux/s390x platforms.
The docker images support ARM architectures which means that they can be used on at least the following versions of Raspberry Pi:
- RPi 1 Model A
- RPi 1 Model A+
- RPi 3 Model A+
- RPi 1 Model B
- RPi 1 Model B+
- RPi 2 Model B
- RPi 2 Model B v1.2 (:heavy_check_mark: tested)
- RPi 3 Model B
- RPi 3 Model B+ (:heavy_check_mark: tested)
- RPi 4 Model B (:heavy_check_mark: tested)
- Compute Module 1
- Compute Module 3
- Compute Module 3 Lite
- Compute Module 3+
- Compute Module 3+ Lite
- RPi Zero PCB v1.2
- RPi Zero PCB v1.3
- RPi Zero W
If you have any of those devices and have successfully used the images feel free to report them and help update this list. 👐
The container by default installs tesseract and 3 datapacks:
tesseract-ocr- English (included)tesseract-ocr-data-deu- Germantesseract-ocr-data-fra- Frenchtesseract-ocr-data-kat- Georgiatesseract-ocr-data-pol- Polishtesseract-ocr-data-rus- Russian
To add more languages you can extend this image and install one or more available language datapacks with the package manager:
FROM hertzg/tesseract-server:latest
RUN apk add --no-cache tesseract-ocr-data-spa tesseract-ocr-data-ara # and so onAfter starting the container the new language will be automatically available.
There are a few endpoints exposed this section describes each one.
This endpoint performs OCR on provided file, You can control the OCR process
by providing options field with JSON object containing the configuration.
This is the main endpoint that expects http multipart request containing
options and file fields and returns a json containing stdout and
stderr of the tesseract process.
The options json object fields directly relate to the CLI options of
tesseract command.
{
"languages": ['eng'], // -l LANG[+LANG] Specify language(s) used for OCR.
"dpi": 300, // --dpi VALUE Specify DPI for input image.
"pageSegmentationMethod": 3, // --psm NUM Specify page segmentation mode.
"ocrEngineMode": 3, // --oem NUM Specify OCR Engine mode.
"tessDataDir": './dir', // --tessdata-dir PATH Specify the location of tessdata path.,
"userPatternsFile": './file', // --user-words PATH Specify the location of user words file.
"userWordsFile": './file', // --user-patterns PATH Specify the location of user patterns file.
"configParams": { // -c VAR=VALUE Set value for config variables.
"VAR": "VALUE", // Note: You can use tesseract --print-parameters to see all available parameters
},
}The returned response has the following shape
{
"exit": {
"code": 0, // Process exit code
"signal": null // Process signal that caused the exit
},
"stderr": "...", // Tesseract Errors and warnings
"stdout": "..." // Tesseract output that contains the result
}# Get worker status
$ curl http://127.0.0.1:8884/statusReturns the pool and their statuses as JSON. When you make OCR request the first pool will be created and then re-used. This endpoint also shows detailed information about each pool including process pids and eviction flags.
{
data: {
processor: {
pools: [
{
args: '-l eng',
resources: [],
status: {
spareResourceCapacity: 2,
size: 0,
available: 0,
borrowed: 0,
pending: 0,
max: 2,
min: 0,
},
},
],
},
},
}Endpoints:
/.well-known/health/healthy/.well-known/health/live/.well-known/health/ready
The difference between liveness and readiness endpoints is the purpose: readiness should be used to denote whether an application is "ready" to receive requests, and liveness should be used to denote whether an application is "live" (vs. in a state where it should be restarted.
The combined health endpoint is designed for cloud technologies, such as Cloud Foundry which only support a single endpoint for both liveness and readiness checking.