# Accessing T-Net in PointNet in Flux3D.jl

**URL:** <https://discourse.julialang.org/t/accessing-t-net-in-pointnet-in-flux3d-jl/87233>\
**Category:** Machine Learning\
**Tags:** flux3d\
**Created:** [September 13, 2022, 8:46pm UTC](https://discourse.julialang.org/t/accessing-t-net-in-pointnet-in-flux3d-jl/87233 "2022-09-13T20:46:08Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![wsshin](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/wsshin/32/360_2.png) [@wsshin](https://discourse.julialang.org/u/wsshin)\
**Post date:** [September 13, 2022, 8:46pm UTC](https://discourse.julialang.org/t/accessing-t-net-in-pointnet-in-flux3d-jl/87233/1 "2022-09-13T20:46:08Z")

</div>

I have a question about `PointNet` in [`Flux3D.jl`](https://github.com/FluxML/Flux3D.jl).

In the original [PointNet paper](https://openaccess.thecvf.com/content_cvpr_2017/papers/Qi_PointNet_Deep_Learning_CVPR_2017_paper.pdf), the author trained the “T-net” to make PointNet invariant w.r.t. rigid transformations of input point clouds (red dashed box below):

 ![Screen Shot 2022-09-13 at 16.44.33](https://global.discourse-cdn.com/julialang/original/3X/5/3/534aa78604e5851892a61b1d0159b28e78f36da2.png)

Is there a way to access this trained T-net from `PointNet` of `Flux3D.jl`?