STream3R

<div align="center"> <h1> STream3R: Scalable Sequential 3D Reconstruction with Causal Transformer (ICLR 2026) </h1> </div> <div align="center"> <h4> <a href="https://nirvanalan.github.io/projects/stream3r" target='_blank'> <img src="https://img.shields.io/badge/🐳-Project%20Page-blue"> </a> <a href="https://arxiv.org/abs/2508.10893" target='_blank'> <img src="https://img.shields.io/badge/arXiv-2508.10893-b31b1b.svg"> </a> <a href="https://arxiv.org/abs/2508.10893" target='_blank'> <img src="https://img.shields.io/badge/OpenreRiew-RTTYGeC2Io-red"> </a> <img src="https://visitor-badge.laobi.icu/badge?page_id=yhluo.STream3R"> </h4> <div > <a href='https://nirvanalan.github.io/' target='_blank'>Yushi Lan</a>^1*&emsp; <a href='https://scholar.google.com/citations?user=fZxK2B0AAAAJ&hl' target='_blank'>Yihang Luo</a>^1*&emsp; <a href='https://hongfz16.github.io' target='_blank'>Fangzhou Hong</a>¹&emsp; <a href='https://shangchenzhou.com/' target='_blank'>Shangchen Zhou</a>¹&emsp; <a href='https://chenhonghua.github.io/clay.github.io/' target='_blank'>Honghua Chen</a>¹&emsp; <br> <a href='https://zhaoyanglyu.github.io/' target='_blank'>Zhaoyang Lyu</a>²&emsp; <a href='https://williamyang1991.github.io/' target='_blank'>Shuai Yang</a>³&emsp; <a href='https://daibo.info/' target='_blank'>Bo Dai</a> ⁴ <a href='https://www.mmlab-ntu.com/person/ccloy/' target='_blank'>Chen Change Loy</a> ¹ &emsp; <a href='https://xingangpan.github.io/' target='_blank'>Xingang Pan</a> ¹ </div> <div> S-Lab, Nanyang Technological University¹; <!-- &emsp; --> <br> Shanghai Artificial Intelligence Laboratory²; WICT, Peking University³; The University of Hong Kong ⁴ <!-- <br> ^*corresponding author --> </div> </div> <br> <div align="center"> <p> <span style="font-variant: small-caps;"><strong>STream3R</strong></span> reformulates dense 3D reconstruction into a sequential registration task with causal attention. <br> <i>⭐ Now supports <b>FlashAttention</b>, <b>KV Cache</b>, <b>Causal Attention</b>, <b>Sliding Window Attention</b>, and <b>Full Attention</b>!</i> </p> <img width="820" alt="pipeline" src="assets/teaser_dynamic.gif"> :open_book: See more visual results on our <a href="https://nirvanalan.github.io/projects/stream3r" target="_blank">project page</a> </div> <br> <details> <summary><b>Abstract</b></summary> <br> <div align="center"> <img width="820" alt="pipeline" src="assets/pipeline.png"> <p align="justify"> We present STream3R, a novel approach to 3D reconstruction that reformulates pointmap prediction as a decoder-only Transformer problem. Existing state-of-the-art methods for multi-view reconstruction either depend on expensive global optimization or rely on simplistic memory mechanisms that scale poorly with sequence length. In contrast, STream3R introduces an streaming framework that processes image sequences efficiently using causal attention, inspired by advances in modern language modeling. By learning geometric priors from large-scale 3D datasets, STream3R generalizes well to diverse and challenging scenarios, including dynamic scenes where traditional methods often fail. Extensive experiments show that our method consistently outperforms prior work across both static and dynamic scene benchmarks. Moreover, STream3R is inherently compatible with LLM-style training infrastructure, enabling efficient large-scale pretraining and fine-tuning for various downstream 3D tasks. Our results underscore the potential of causal Transformer models for online 3D perception, paving the way for real-time 3D understanding in streaming environments. </p> </div> </details>

:fire: News

• [Mar 9, 2026] Check out the DUSt3R-based metric-scale STream3R[α] version on the alpha branch.
• [Jan 26, 2026] Accepted to ICLR 2026!
• [Sep 16, 2025] The complete training code is released!
• [Aug 22, 2025] The evaluation code is now available!
• [Aug 15, 2025] Our inference code and weights are released!

🔧 Installation

1. Clone Repo

   git clone https://github.com/NIRVANALAN/STream3R
   cd STream3R
   

2. Create Conda Environment

   conda create -n stream3r python=3.11 cmake=3.14.0 -y
   conda activate stream3r
   

3. Install Python Dependencies

   Important: Install Torch based on your CUDA version. For example, for Torch 2.8.0 + CUDA 12.6:

   # Install Torch
   pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu126
   
   # Install other dependencies
   pip install -r requirements.txt
   
   # Install STream3R as a package
   pip install -e .
   

:computer: Inference

You can now try STream3R with the following code. The checkpoint will be downloaded automatically from Hugging Face.

You can set the inference mode to causal for causal attention, window for sliding window attention (with a default window size of 5), or full for bidirectional attention.

import os
import torch
from stream3r.models.stream3r import STream3R
from stream3r.models.components.utils.load_fn import load_and_preprocess_images

device = "cuda" if torch.cuda.is_available() else "cpu"

model = STream3R.from_pretrained("yslan/STream3R").to(device)
model.eval()

example_dir = "examples/static_room"
image_names = [os.path.join(example_dir, file) for file in sorted(os.listdir(example_dir))]
images = load_and_preprocess_images(image_names).to(device)

with torch.no_grad():
    # Use one mode "causal", "window", or "full" in a single forward pass
    predictions = model(images, mode="causal")

A script is already at inference_stream3r.py.

We also support a KV cache version to enable streaming input using StreamSession. The StreamSession takes sequential input and processes them one by one, making it suitable for real-time or low-latency applications. This streaming 3D reconstruction pipeline can be applied in various scenarios such as real-time robotics, autonomous navigation, online 3D understanding and SLAM. An example usage is shown below:

import os
import torch
from stream3r.models.stream3r import STream3R
from stream3r.stream_session import StreamSession
from stream3r.models.components.utils.load_fn import load_and_preprocess_images

device = "cuda" if torch.cuda.is_available() else "cpu"

model = STream3R.from_pretrained("yslan/STream3R").to(device)

example_dir = "examples/static_room"
image_names = [os.path.join(example_dir, file) for file in sorted(os.listdir(example_dir))]
images = load_and_preprocess_images(image_names).to(device)
# StreamSession supports KV cache management for both "causal" and "window" modes.
session = StreamSession(model, mode="causal")

with torch.no_grad():
    # Process images one by one to simulate streaming inference
    for i in range(images.shape[0]):
        image = images[i : i + 1]
        predictions = session.forward_stream(image)
    session.clear()

:zap: Demo

You can run the demo built on VGG-T's code using the script app.py with the following command:

python app.py

📁 Code Structure

The repository is structured as follows:

STream3R/
├── stream3r/                    
│   ├── models/                  
│   │   ├── stream3r.py            
│   │   ├── multiview_dust3r_module.py  
│   │   └── components/               
│   ├── dust3r/                 
│   ├── croco/                  
│   ├── utils/                  
│   └── stream_session.py          
├── configs/                     
├── examples/                    
├── assets/                      
├── app.py                          
├── requirements.txt                 
├── setup.py                        
└── README.md                       

:100: Quantitive Results

3D Reconstruction Comparison on NRGBD.

| Method | Type | Acc Mean ↓ | Acc Med. ↓ | Comp Mean ↓ | Comp Med. ↓ | NC Mean ↑ | NC Med. ↑ | |---------------------|----------|------------|------------|-------------|-------------|-----------|-----------| | VGG-T | FA | 0.073 | 0.018 | 0.077 | 0.021 | 0.910 | 0.990 | | DUSt3R | Optim | 0.144 | 0.019 | 0.154 | 0.018 | 0.870 | 0.982 | | MASt3R | Optim | 0.085 | 0.033 | 0.063 | 0.028 | 0.794 | 0.928 | | MonST3R | Optim | 0.272 | 0.114 | 0.287 | 0.110 | 0.758 | 0.843 | | Spann3R | Stream | 0.416 | 0.323 | 0.417 | 0.285 | 0.684 | 0.789 | | CUT3R | Stream | 0.099 | 0.031 | 0.076 | 0.026 | 0.837 | 0.971 | | StreamVGGT | Stream | 0.084 | 0.044 | 0.074 | 0.041 | 0.861 | 0.986 | | Ours | Stream | 0.057 | 0.014 | 0.028 | 0.013 | 0.910 | 0.993 |

Read our full paper for more insights.

⏳ GPU Memory Usage and Runtime

We report the peak GPU memory usage (VRAM) and runtime of our full model for processing each streaming input using the StreamSession implementation. All experiments were conducted at a common resolution of 518 × 384 on a single H200 GPU. The benchmark includes both Causal for causal attention and Window for sliding window attention with a window size of 5.

*Run Time (s)