# Memory Resident Vision Model

**URL:** <https://discuss.ai.google.dev/t/memory-resident-vision-model/30182>\
**Category:** General Discussion\
**Tags:** tfimage, pytorch, model\_maker\
**Created:** [March 7, 2024, 2:39am UTC](https://discuss.ai.google.dev/t/memory-resident-vision-model/30182 "2024-03-07T02:39:42Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![bruce\_baur](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/bruce_baur/32/16367_2.png) [@bruce\_baur](https://discuss.ai.google.dev/u/bruce_baur)\
**Post date:** [March 7, 2024, 2:39am UTC](https://discuss.ai.google.dev/t/memory-resident-vision-model/30182/1 "2024-03-07T02:39:43Z")

</div>

I have a pytorch model working using computer vision and each time I need to run detect.py. This takes a few seconds to load and produce the output predictions. I was seeking a way to have a CV model always running and ready to receive new images for it to detect objects on. What do you recommend in order to have a memory resident model always on, such that I can continue sending individual images and not have to reload the parameters into memory each and every time?