Neural Slicer for Multi-Axis 3D Printing (SIGGRAPH 2024)
Vložit
- čas přidán 17. 06. 2024
- In this work, we employ neural networks to establish a deformation mapping for curved layer generation that can be optimized through loss functions directly defined on the local printing directions. Our method relies less on the initial values and can generate results with significantly improved performance for models in diverse representations and topology.
This research work has been accepted to publish in the following paper:
Tao Liu, Tianyu Zhang, Yongxue Chen, Yuming Huang, and Charlie C.L. Wang, "Neural slicer for multi-axis 3D printing", ACM Transactions on Graphics (SIGGRAPH 2024).
Project Page: ryantaoliu.github.io/NeuralSl...
(Source code will be released after July 2024) - Věda a technologie
Amazing Work! i dont doubt we will start seeing machines with more than 3 axis as the new normal! thank you
Thanks. Source code will be released on the project page after July 2024. As required by ACM SIGGRAPH, we cannot release it before the conference.
Incredible work as always, pushing the limits!
Looks incredible !
This could be very useful for printing stronger prosthetics while still keeping the easily adjustable and fast method of FDM printers. Very cool to look into
Wow, I didn't know such 3D printers existed!
Is it possible to limit the slope of the layers relative to a horizontal stationary bed to make it useful for hobbyist off the shelf printers?
Yes it can be realized by controlling the allowed local printing directions in the model space.
@@CharlieCLWang-od4px hyper dimensional bed slinging
This is incredible. Eagerly awaiting code so we can replicate!
Amazing progress, lots going on at Manchester
@CNCKitchen @TeachingTech
@NathanBuildsRobots
Congratulation!
So cool ! No support 🤩
Very cool! Are there plans to streamline the algorithm for third party tool usage? Doing so would allow it to separate itself from current non-planar slicers in the sense that they are "gimmick" tools. Also, are there plans to open source or source available the project?
when i read neural slicer i imagined something like neuralink
This is amazing! What are the current limitations you've found?
Nevermind, I find it on the GitHub page
Will the idex style hotbeds be able to use this? Specifically I am talking about where they use the 4 corners and individually adjust them to change the angle of the bed. Rn they only use it for bed leveling, but this software could enable a lot for these printers, right?
В рф такой давно в комерции)
Are there other advantages besides being more capable at handeling compelx designs?
Who wants such jagged meshes? is this low quality just to show speed?
Neat but the print quality is awful. That really needs to improve.
You are welcome to improve it yourself clown
Because today everything has to be Neural something even though it makes no sense. Well sounds less like Neural Network is that much important here but yeh funding these days..
Reason: We employ NN to represent meshfree continuous functions for deformation so that can borrow the computational power of the modern machine learning pipeline such as the robust stochastic gradient solver and parallel computing to solve this highly nonlinear problem in a way much better than our earlier approach in S^3-slicer (video: czcams.com/video/qNm1ierKuUk/video.html).
Sit down before you hurt yourself