Cast category_id field value from str to int

[dnth]

This PR casts the category_id field value from str to int data type to be consistent with the COCO data format.

I have also checked the other fields and found everything is consistent with the format outlined in https://cocodataset.org/#format-data

Closes #46

[alexheat]

Hi @dnth , thank you! But it looks like the change you made has failed one of the test cases. It is getting an errror " ValueError: cannot convert float NaN to integer"

That test is running this notebook https://github.com/pylabel-project/samples/blob/main/yolo2coco.ipynb

Could you try your change on the dataset that is used in that notebook? I think there may be some null values that are causing the error. Most of the issues that people report are due to some kind of null values in the data sets.

There are safe ways that you can convert the field to an int while leaving null values as null, and making an error.

You can see this readme for how to run the test cases before you make a pull request https://github.com/pylabel-project/pylabel/tree/dev/tests

Thanks again Dickson

[alexheat]

This conversion throws an error if the cat_id is null.

[dnth]

Hi @dnth , thank you! But it looks like the change you made has failed one of the test cases. It is getting an errror " ValueError: cannot convert float NaN to integer"

That test is running this notebook https://github.com/pylabel-project/samples/blob/main/yolo2coco.ipynb

Could you try your change on the dataset that is used in that notebook? I think there may be some null values that are causing the error. Most of the issues that people report are due to some kind of null values in the data sets.

There are safe ways that you can convert the field to an int while leaving null values as null, and making an error.

You can see this readme for how to run the test cases before you make a pull request https://github.com/pylabel-project/pylabel/tree/dev/tests

Thanks again Dickson

Ahh i see.. I didn't consider that before I made the PR.. I have amended to check for na value and ran the tests. Tests are passing on my machine. Could you please check if the code is optimal?

[dnth]

I find that I need to install the following before I can run the test:

pip install pytest
pip install pytest-lazy-fixture
pip install nbmake

Do you think we should add it in the test readme to make it easier for first time contributors?

[alexheat]

Thank you @dnth that looks like a good fix. I have accepted the PR and release v.37 with the fix. Could you give it a try?

Also I have added some info to the tests readme on how to install the requirements https://github.com/pylabel-project/pylabel/tree/dev/tests.

And I should also make a better guide for contributors. That is on my to do list :)

[dnth]

Thank you @dnth that looks like a good fix. I have accepted the PR and release v.37 with the fix. Could you give it a try?

Also I have added some info to the tests readme on how to install the requirements https://github.com/pylabel-project/pylabel/tree/dev/tests.

And I should also make a better guide for contributors. That is on my to do list :)

Yay! Looks good to me!