[OCR] Convert jupyter notebook (with bash script) to Python script

[cuducos]

Currently the code that generates a dataset with the text from CEAP receipts is in this notebook.

As it uses shell scripts here and there it would be great to have this as a standard src/ Python file, without shell script, to automate this data collection without some many dependencies such as binaries available in one's $PATH.

[jtemporal]

#207 documents how to generate the dataset that already is on our S3 but there's no python script for it. thanks @fgrehm for the data btw ;)

[tuliocasagrande]

Hello @cuducos and @jtemporal

I took a quick look at the problem and it seems we're pending on the ReimbursementOCR to be implemented. Is that correct?
The google-cloud-vision python library can also be a good idea.

[cuducos]

I took a quick look at the problem and it seems we're pending on the ReimbursementOCR to be implemented. Is that correct?

Exactly — a bit of background just in case ; )

[fgrehm]

I actually did the OCR with python, it even run things in parallel 😄 :mindblown:

Just check the stuff I linked on #188 (comment) and LMK if u need any help with that!

[jandersoncoelho]

Have you seen http://ocrmypdf.readthedocs.io/ ? I use that in my research. It's a interface to Tesseract-ocr and the results It seems good.

[fgrehm]

Heads up: I've been hacking away on a better approach for OCRing receipts at https://github.com/fgrehm/serenata-ocr and one of the ideas is that it will have support for a "pluggable provider interface", meaning people can choose between Google Cloud Vision, https://ocr.space/, Microsoft Azure and maybe even some self hosted tesseract infra.

[fgrehm]

See #298