# Deploying action recognition model

**URL:** <https://discuss.pytorch.org/t/deploying-action-recognition-model/88713>\
**Category:** deployment\
**Created:** [July 10, 2020, 5:20pm UTC](https://discuss.pytorch.org/t/deploying-action-recognition-model/88713 "2020-07-10T17:20:28Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![braindotai](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/braindotai/32/61373_2.png) [@braindotai](https://discuss.pytorch.org/u/braindotai)\
**Post date:** [July 10, 2020, 5:20pm UTC](https://discuss.pytorch.org/t/deploying-action-recognition-model/88713/1 "2020-07-10T17:20:28Z")

</div>

If performance is the priority here, would it be okay to deploy a heavy model for action recognition natively on mobile (using pytorch mobile)?

Or should I go with something like rest api, hosted on other remote server (amazon, azure, etc).

---

<div class="post-metadata">

**Author:** ![Shisho\_Sama](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/shisho_sama/32/6926_2.png) [@Shisho\_Sama](https://discuss.pytorch.org/u/Shisho_Sama)\
**Post date:** [July 11, 2020, 9:24am UTC](https://discuss.pytorch.org/t/deploying-action-recognition-model/88713/2 "2020-07-11T09:24:03Z")

</div>

Having it on remote is more logical unless you have very bad internet connection.  
its not customary to place the actual model on the device, as it puts everything into risk.