# Ros\_ncnn: highly optimized neural inference framework

**URL:** <https://discourse.openrobotics.org/t/ros-ncnn-highly-optimized-neural-inference-framework/14636>\
**Category:** Projects\
**Created:** [June 12, 2020, 11:26am UTC](https://discourse.openrobotics.org/t/ros-ncnn-highly-optimized-neural-inference-framework/14636 "2020-06-12T11:26:09Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![nilseuropa](https://sea2.discourse-cdn.com/flex022/user_avatar/discourse.openrobotics.org/nilseuropa/32/6770_2.png) [@nilseuropa](https://discourse.openrobotics.org/u/nilseuropa)\
**Post date:** [June 12, 2020, 11:26am UTC](https://discourse.openrobotics.org/t/ros-ncnn-highly-optimized-neural-inference-framework/14636/1 "2020-06-12T11:26:09Z")

</div>

Hi,

I couldn’t find a wrapper for a fast enough inference engine, so here is one for `ncnn`:  
[https://github.com/nilseuropa/ros\_ncnn](https://github.com/nilseuropa/ros_ncnn)

It features several popular neural models ( e.g. yocto, yolact, retinaface ), I am also planning on extending the model library with non computer vision models as well.

cheers,  
m.