# External data and the use of an ImageNet pre-trained model

**URL:** <https://community.drivendata.org/t/external-data-and-the-use-of-an-imagenet-pre-trained-model/3682>\
**Category:** Mapping Disaster Risk from Aerial Imagery\
**Created:** [October 3, 2019, 5:22pm UTC](https://community.drivendata.org/t/external-data-and-the-use-of-an-imagenet-pre-trained-model/3682 "2019-10-03T17:22:36Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![sallamander](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/sallamander/32/841_2.png) [@sallamander](https://community.drivendata.org/u/sallamander)\
**Post date:** [October 3, 2019, 5:22pm UTC](https://community.drivendata.org/t/external-data-and-the-use-of-an-imagenet-pre-trained-model/3682/1 "2019-10-03T17:22:36Z")

</div>

Is the use of an ImageNet pre-trained model as shown in the [benchmark code](http://drivendata.co/blog/disaster-response-roof-type-benchmark/) that is linked in the [problem description](https://www.drivendata.org/competitions/58/disaster-response-roof-type/page/143/) page a use of external data, and therefore against the rules, or does that not count?

---

<div class="post-metadata">

**Author:** ![sallamander](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/sallamander/32/841_2.png) [@sallamander](https://community.drivendata.org/u/sallamander)\
**Post date:** [October 9, 2019, 11:50am UTC](https://community.drivendata.org/t/external-data-and-the-use-of-an-imagenet-pre-trained-model/3682/2 "2019-10-09T11:50:39Z")

</div>

Is there an administrator or somebody at Driven Data that can potentially answer this question… @bull, maybe?

---

<div class="post-metadata">

**Author:** ![nrw09](https://avatars.discourse-cdn.com/v4/letter/n/e19b73/32.png) [@nrw09](https://community.drivendata.org/u/nrw09)\
**Post date:** [October 9, 2019, 12:59pm UTC](https://community.drivendata.org/t/external-data-and-the-use-of-an-imagenet-pre-trained-model/3682/3 "2019-10-09T12:59:51Z")

</div>

I have a related question: what about models that were pre-trained on SpaceNet data, such as those in [solaris](https://solaris.readthedocs.io/en/latest/)?

---

<div class="post-metadata">

**Author:** ![florianm](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/florianm/32/846_2.png) [@florianm](https://community.drivendata.org/u/florianm)\
**Post date:** [October 14, 2019, 9:11pm UTC](https://community.drivendata.org/t/external-data-and-the-use-of-an-imagenet-pre-trained-model/3682/4 "2019-10-14T21:11:29Z")

</div>

Any updates on this?

---

<div class="post-metadata">

**Author:** ![glipstein](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/glipstein/32/920_2.png) [@glipstein](https://community.drivendata.org/u/glipstein)\
**Post date:** [October 15, 2019, 6:12pm UTC](https://community.drivendata.org/t/external-data-and-the-use-of-an-imagenet-pre-trained-model/3682/5 "2019-10-15T18:12:52Z")

</div>

Hi all, thanks for these questions! We just added a post that covers the answers:

> [@Official external data / pre-trained models thread](http://community.drivendata.org/t/official-external-data-pre-trained-models-thread/3740):
>
> Pre-trained models in particular are allowed as long as they can be released under an [Open Source License](https://opensource.org/licenses). If you do use a pre-trained model, please make sure to share in this thread for the rest of the community. Thanks and good luck!

Enjoy and good luck!
