Skip to content

Inference in OCR is returning the wrong text. #2

@lucas-g-cardoso

Description

@lucas-g-cardoso

I am trying to run inference in the PlotQA pipeline, and in the OCR stage, I am trying to transform the detection files (in the format of CLASS_LABEL CLASS_CONFIDENCE XMIN YMIN XMAX YMAX) into .csv files, using the scripts given in the pipeline. The main script is the ocr_and_sie.py. The problem is that this script returns wrong texts.

After trying to underestand the problem, I found out that the doOCR function (in ocr_and_sie.py file) is responsible for returning wrong strings to put in the CSV, although i don't know why.

Here are some examples of wrong text it returns: )ear, RVHE’IJT‘ 1‘1 and DPUIIIi'pt‘fiwtfl, and so on...

If anyone could help me with this issue I would be very thankful.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions