---
license: mit
---
# Mamba3D 

[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/mamba3d-enhancing-local-features-for-3d-point/supervised-only-3d-point-cloud-classification)](https://paperswithcode.com/sota/supervised-only-3d-point-cloud-classification?p=mamba3d-enhancing-local-features-for-3d-point)
[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/mamba3d-enhancing-local-features-for-3d-point/3d-point-cloud-classification-on-modelnet40)](https://paperswithcode.com/sota/3d-point-cloud-classification-on-modelnet40?p=mamba3d-enhancing-local-features-for-3d-point)
[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/mamba3d-enhancing-local-features-for-3d-point/3d-point-cloud-classification-on-scanobjectnn)](https://paperswithcode.com/sota/3d-point-cloud-classification-on-scanobjectnn?p=mamba3d-enhancing-local-features-for-3d-point)

This repository contains the official implementation of the paper:

[**[ACM MM 24] Mamba3D: Enhancing Local Features for 3D Point Cloud Analysis via State Space Model**](https://arxiv.org/abs/2404.14966)


## 📰 News
- [2024/8] We release the training and evaluation code [here](https://github.com/xhanxu/Mamba3D)! Pretrained weights are [here](./pretrain_pointmae/ckpt-last.pth)!
- [2024/7] Our [MiniGPT-3D](https://github.com/tangyuan96/minigpt-3d) is also accepted by ACM MM24! We outperform existing large point-language models, using just about 1 day on 1 RTX 3090! Check it out!
- [2024/7] Ours Mamba3D is accepted by ACM MM24!

- [2024/4] We present Mamba3D, a state space model tailored for point cloud learning.
<div style="text-align: center">
<img src="media/mamba3d_total_v2.png" />
</div>

# 🥶 Code: https://github.com/xhanxu/Mamba3D

# 🐟 Pretrained weights:

You can find the pre-trained weights [here](./pretrain_pointmae/ckpt-last.pth).

Or, specifically as follows.
| Dataset             | Pretrain | Acc   | Weight |
|---------------------|----------|-------|--------|
|ShapeNet             | Point-MAE|    |[ckpt](./pretrain_pointmae/ckpt-last.pth)   |  
| ModelNet40          | no       | 93.4  | [ckpt](./modelnet40_scratch_93.4/ckpt-best.pth)    |
| ModelNet40          | Point-MAE | 94.7  | [ckpt](./modelnet40_pointmae_94.7/ckpt-best.pth)    |
| ScanObjectNN-hardest| no       | 91.81 | [ckpt](./scanobjectnn_hardest_scratch_91.81/ckpt-best-91.8.pth)    |
| ScanObjectNN-hardest| Point-MAE | 92.05 | [ckpt](./scanobjectnn_hardest_pointmae_92.05/ckpt-best.pth)    |

## 😀 Contact
If you have any questions or are looking for cooperation in related fields, please contact [Xu Han](https://xhanxu.github.io/) via xhanxu@hust.edu.cn. 


---
license: mit
---