Error opening data file /usr/share/tessdata/.traineddata

[jens1o]

Following example code:

extern crate tesseract;
use tesseract::Tesseract;

fn main() {
    let tesseract_instance = Tesseract::new();

    tesseract_instance.set_lang("deu");
    tesseract_instance.set_image("sample.jpg");
    dbg!(tesseract_instance.get_text());
}

The problem is that I do have tesseract training data installed, but it is named {language_name}.traineddata(for example: deu.traineddata), so this wrapper can't find it. When I manually run tesseract sample.jpg test.txt -l deu, it works seamlessly.

How can I manually change the file naming scheme?

[guygastineau]

I have this same issue. I just read the source for this crate on crates.io but I couldn't find where it is set.

[ccouzens]

Hello,

There is an environment variable TESSDATA_PREFIX. Would that help?

There is also a datapath parameter we could set inside set_lang. It's currently set to ptr::null(). I'm looking into exposing that as part of the API. Would that help?

This crate is a wrapper around Tesseract's c api. Do any of the functions there look like they might help? If so, I'll expose them.

[guygastineau]

Thanks for the reply, I'll mess with your suggestions and see what happens.

[GirkovArpa]

The environment variable TESSDATA_PREFIX can be set to the executable's directory like this:

// build.rs in top level folder (not in src folder)
fn main() {
    // load testdata file from same folder as executable
    println!("cargo:rustc-env=TESSDATA_PREFIX=");
}