# How to use compute metrics

**URL:** <https://community.drivendata.org/t/how-to-use-compute-metrics/6350>\
**Category:** Image Similarity Challenge\
**Created:** [July 29, 2021, 5:27pm UTC](https://community.drivendata.org/t/how-to-use-compute-metrics/6350 "2021-07-29T17:27:41Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![dfrid](https://avatars.discourse-cdn.com/v4/letter/d/e5b9ba/32.png) [@dfrid](https://community.drivendata.org/u/dfrid)\
**Post date:** [July 29, 2021, 5:27pm UTC](https://community.drivendata.org/t/how-to-use-compute-metrics/6350/1 "2021-07-29T17:27:41Z")

</div>

Lets say I have a function that takes 2 images and returns their similarities. How can I use this with compute metrics to evaluate my model?

---

<div class="post-metadata">

**Author:** ![jayqi](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/jayqi/32/924_2.png) [@jayqi](https://community.drivendata.org/u/jayqi)\
**Post date:** [July 29, 2021, 7:24pm UTC](https://community.drivendata.org/t/how-to-use-compute-metrics/6350/2 "2021-07-29T19:24:41Z")

</div>

Hi @dfrid. The `eval_metrics.py` script provided on the [Data Download page](https://www.drivendata.org/competitions/80/competition-image-similarity-2-dev/data/) takes input files conforming to the competition submission formats from either the Matching Track or the Descriptor Track. You can read more about the submission formats on the respective Problem Description pages:

- [Matching Track Submission Format](https://www.drivendata.org/competitions/79/competition-image-similarity-1-dev/page/376/#submissionformat)
- [Descriptor Track Submission Format](https://www.drivendata.org/competitions/80/competition-image-similarity-2-dev/page/380/#submissionformat)