ICDAR dataset question

I want to test the accuracy of text-recognition, text-detection and text-recognition.

And I download the ICARD dataset from here.

When I run the command: accuracy_check -c models/intel/text-recognition-0012/accuracy-check.yml -s /home/g/xiaowei/dataset/ICDAR/Word-Recognition-2013-edition -m intel/text-recognition-0012/FP16 -d tools/accuracy_checker/dataset_definitions.yml -td CPU

I meet the error:

Processing info:
model: text-recognition-0012
launcher: dlsdk
device: CPU
dataset: ICDAR2013
OpenCV version: 4.5.1-openvino
Annotation conversion for ICDAR2013 dataset has been started
Parameters to be used for conversion:
converter: icdar13_recognition
annotation_file: /home/g/xiaowei/dataset/ICDAR/Word-Recognition-2013-edition/ICDAR13_REC/gt/gt.txt.fixed.alfanumeric
10:41:29 accuracy_checker ERROR: Invalid value "/home/g/xiaowei/dataset/ICDAR/Word-Recognition-2013-edition/ICDAR13_REC/gt/gt.txt.fixed.alfanumeric" for icdar13_recognition_converter_config.annotation_file: path does not exist

The dataset.md have no more information for this dataset. And I don’t know how to get gt/gt.txt.fixed.alfanumeric.

Can you share more information for ICDAR dataset?

Thank you!

[WenjingKangIntel]

Hi. You should mkdir /home/g/xiaowei/dataset/ICDAR/Word-Recognition-2013-edition/ICDAR13_REC/gt/ and put Challenge2_Test_Task3_GT.txt there and rename Challenge2_Test_Task3_GT.txt to gt.txt.fixed.alfanumeric.

[xiaoweiChen]

Thanks you reply. As your words, I know:

gt.txt.fixed.alfanumeric is a subset of icdar13 test dataset.

This is important for me, and I know the reason why not match 0.8818.

For this line word_24.png, "[falling , I remove all words, it become a empty line. word_24.png, "[falling -> And then, I get 87.94%.

You give me the right answer for this issue, I think this issue should be closed.

Thanks again for @cocokale and @morkovka1337 ：）