# Image Classification with the ability to verify Fonts and image abnormalities or errors in the image

**URL:** <https://discuss.ai.google.dev/t/image-classification-with-the-ability-to-verify-fonts-and-image-abnormalities-or-errors-in-the-image/28497>\
**Category:** Keras\
**Tags:** keras, models, help\_request\
**Created:** [November 23, 2022, 6:50am UTC](https://discuss.ai.google.dev/t/image-classification-with-the-ability-to-verify-fonts-and-image-abnormalities-or-errors-in-the-image/28497 "2022-11-23T06:50:56Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Rod\_Neufeld](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/rod_neufeld/32/12976_2.png) [@Rod\_Neufeld](https://discuss.ai.google.dev/u/Rod_Neufeld)\
**Post date:** [November 23, 2022, 6:50am UTC](https://discuss.ai.google.dev/t/image-classification-with-the-ability-to-verify-fonts-and-image-abnormalities-or-errors-in-the-image/28497/1 "2022-11-23T06:50:56Z")

</div>

Hi I am working on a project to classify football (Soccer) team crests, I have a working model that accomplishes the initial classifications. I need to develop not only the ability to verify that the image is indeed a particular team crest I need to be able to find errors or possible errors (imperfections) in the image design itself and or font issues that deviate from the correct font for the Team Crest.

Would this kind of verification or sub-classification have to be done post the initial team classification with another model that classifies errors in rendering or font errors?

Thoughts or direction on this is appreciated.

---

<div class="post-metadata">

**Author:** ![LK\_Kadali](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/lk_kadali/32/13996_2.png) [@LK\_Kadali](https://discuss.ai.google.dev/u/LK_Kadali)\
**Post date:** [September 17, 2024, 8:56am UTC](https://discuss.ai.google.dev/t/image-classification-with-the-ability-to-verify-fonts-and-image-abnormalities-or-errors-in-the-image/28497/3 "2024-09-17T08:56:47Z")

</div>

Hi @Rod_Neufeld,

For your use case, you need to train a separate anomaly classification model in finding the imperfections in image design and font errors. To train the model, it requires the new data with correct font and imperfections . Then integrate this model with your team crest classification model.

Thank You