# About Polimago Classification and Regression Project Testing Image

**URL:** <https://forum.commonvisionblox.com/t/about-polimago-classification-and-regression-project-testing-image/855>\
**Category:** Tools\
**Tags:** polimago\
**Created:** [August 19, 2019, 12:07pm UTC](https://forum.commonvisionblox.com/t/about-polimago-classification-and-regression-project-testing-image/855 "2019-08-19T12:07:55Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![fluber](https://avatars.discourse-cdn.com/v4/letter/f/cc9497/32.png) [@fluber](https://forum.commonvisionblox.com/u/fluber)\
**Post date:** [August 19, 2019, 12:07pm UTC](https://forum.commonvisionblox.com/t/about-polimago-classification-and-regression-project-testing-image/855/1 "2019-08-19T12:07:55Z")

</div>

Image Resolution 2048 X 2200 (Using Line Scan Camera Grab Image)  
Image Format MONO  
Image Count 82  
I Create Two Class.  
Class Finger, Sample Count 50  
Class Scratch, Sample Count 163  
The Testing Result

 ![G](https://canada1.discourse-cdn.com/flex027/uploads/commonvisionblox/original/1X/36512fbb426eebda3e41c39bb1d7775017d877d3.png)

I found the test result is building on the images from building classification file. I think should be use a different images for test…

Can i use new images to test classification result, not use the sample images from building classification file ?

---

<div class="post-metadata">

**Author:** ![Frank](https://yyz1.discourse-cdn.com/flex027/user_avatar/forum.commonvisionblox.com/frank/32/42_2.png) [@Frank](https://forum.commonvisionblox.com/u/Frank)\
**Post date:** [August 19, 2019, 1:12pm UTC](https://forum.commonvisionblox.com/t/about-polimago-classification-and-regression-project-testing-image/855/2 "2019-08-19T13:12:46Z")

</div>

> [@fluber](#):
>
> I found the test result is building on the images from building classification file. I think should be use a different images for test…

If you ran the Leave-Out Test in the TeachBench, the test is not performed on images that have been used for training. With the k-fold cross-validation (also called hold-out test or leave-out test), the test is performed on images that have been excluded during training to reflect the classifiers performance on “unknown/new” images.

This is also explained in the mouse-over tool tip:

![grafik](https://canada1.discourse-cdn.com/flex027/uploads/commonvisionblox/original/1X/409c74ec4b81c0bda420b611c7f8b158692efd24.png)