# Transformation on Training Data

**URL:** <https://community.drivendata.org/t/transformation-on-training-data/6354>\
**Category:** Image Similarity Challenge\
**Created:** [July 31, 2021, 2:28pm UTC](https://community.drivendata.org/t/transformation-on-training-data/6354 "2021-07-31T14:28:03Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Vinh\_Tuan\_La](https://avatars.discourse-cdn.com/v4/letter/v/977dab/32.png) [@Vinh\_Tuan\_La](https://community.drivendata.org/u/Vinh_Tuan_La)\
**Post date:** [July 31, 2021, 2:28pm UTC](https://community.drivendata.org/t/transformation-on-training-data/6354/1 "2021-07-31T14:28:03Z")

</div>

At Transformations subsection on the original paper, I notice there are GIMP and AugLy tools to alter images. I wonder when I use the training method, I must transform/alter image myself for training right? Or training data already prepare all manipulate and original images for training?  
I also wonder is there any GIMP transformation package for python, it is convenient if it’s exist.

---

<div class="post-metadata">

**Author:** ![mike-dd](https://avatars.discourse-cdn.com/v4/letter/m/f0a364/32.png) [@mike-dd](https://community.drivendata.org/u/mike-dd)\
**Post date:** [August 2, 2021, 4:27pm UTC](https://community.drivendata.org/t/transformation-on-training-data/6354/2 "2021-08-02T16:27:34Z")

</div>

Hi Vinh\_Tuan\_La, welcome to the competition!

The “training images” are not transformed, so yes, you will need to apply these transformations yourself.

You may find it helpful to read the [Getting Started blog post](https://www.drivendata.co/blog/image-similarity-challenge/) which provides some background on the intended use of the training images.

Good luck!