A High-Performance Distributed Library for Large-Scale Bundle Adjustment

Overview

MegBA: A High-Performance and Distributed Library for Large-Scale Bundle Adjustment

This repo contains an official implementation of MegBA.

MegBA is a fast and distributed library for large-scale Bundle Adjustment (BA). MegBA has a novel end-to-end vectorised BA algorithm which can fully exploit the massive parallel cores on GPUs, thus speeding up the entire BA computation. It also has a novel distributed BA algorithm that can automatically partition BA problems, and solve BA sub-problems using distributed GPUs. The GPUs synchronise intermediate solving state using network-efficient collective communication, and the synchronisation is designed to minimise communication cost. MegBA has a memory-efficient GPU runtime and it exposes g2o-compatible APIs. Experiments show that MegBA can out-perform state-of-the-art BA libraries (i.e., Ceres and DeepLM) by ~50x and ~5x respectively, in public large-scale BA benchmarks.

Version

  • 2021/12/06 Beta version released! It corresponds to this paper.
  • General version code release (Expected Dec 31 2021)
  • memory-efficient version with implicit Hessian (TBD)
  • analytical differential module, IMU factor, prior factor (TBD)

Paper:

Quickstart

Dependencies:

You can also easily install all dependencies with script: script

Demo with BAL dataset:

  • Download any pre.txt.bz2 file from BAL Dataset: https://grail.cs.washington.edu/projects/bal/ and uncompressed.

  • Compile

    mkdir build
    cd build
    cmake ..
    make -j4 BAL_Double
  • Run the demo (Venice-1778)

    cd examples
    ./BAL_Double --name=Venice --world_size=2 --iter=100 --solver_tol=1e-1 --solver_refuse_ratio=1 --solver_max_iter=100 --tau=1e4 --epsilon1=1 --epsilon2=1e-10
    • world_size: number of GPUs available
    • iter: the maximal number of LM iteration
    • epsilon: threshold in LM
    • solver_tol: tolerance of solver (distributed PCG solver)
    • solver_refuse_ratio: early stop for the solver
    • solver_max_iter: the maximal iteration of solver
    • tau: the initial region

Notes for the practitioners

  • Currently, MegBA implements automatic differentation only for generalizability. Please consider implementing your own analytical differentiation module.
  • If you use devices without modern inter-device communication (i.e., NVLinks..), you might find the data transfer is the bottleneck.
  • Empirically, we found it is necessary to customize the LM trust-region strategies and tune its hyper-parameters to further boost the performance.

Documentation

Under doc/ (Coming soon...)

Collaborate with Us

Please check here for MegBA's future plan.

If you are intereted in MegBA and want to collaborate, you can:

  • Apply for an Internship at Megvii Research 3D, please send your resume to [email protected], with your expected starting date. (subject: 3D组CUDA实习生-Name) Unfortunately, now we are only able to host interns with work authorization in China.
  • As an external collaborator (coding), just fork this repo and send PRs. We will review your PR carefully (and merge it into MegBA).
  • As an algorithm/novelty contributor, please send an email to [email protected].
  • Any new feature request, you can send an email to [email protected] as well. Note that it is not guaranteed the requested feature will be added or added soon

Contact Information:

BibTeX Citation

If you find MegBA useful for your project, please consider citing:

@misc{2021megba,
  title={MegBA: A High-Performance and Distributed Library for Large-Scale Bundle Adjustment}, 
  author={Jie Ren and Wenteng Liang and Ran Yan and Luo Mai and Shiwen Liu and Xiao Liu},
  year={2021},
  eprint={2112.01349},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}

License

MegBA is licensed under the Apache License, Version 2.0.

Owner
旷视研究院 3D 组
旷视科技(Face++)研究院 3D 组(原 SLAM 组)
旷视研究院 3D 组
Portfolio asset allocation strategies: from Markowitz to RNNs

Portfolio asset allocation strategies: from Markowitz to RNNs Research project to explore different approaches for optimal portfolio allocation starti

Luigi Filippo Chiara 1 Feb 05, 2022
Unsupervised Image Generation with Infinite Generative Adversarial Networks

Unsupervised Image Generation with Infinite Generative Adversarial Networks Here is the implementation of MICGANs using DCGAN architecture on MNIST da

16 Dec 24, 2021
Unsupervised phone and word segmentation using dynamic programming on self-supervised VQ features.

Unsupervised Phone and Word Segmentation using Vector-Quantized Neural Networks Overview Unsupervised phone and word segmentation on speech data is pe

Herman Kamper 13 Dec 11, 2022
Distance Encoding for GNN Design

Distance-encoding for GNN design This repository is the official PyTorch implementation of the DEGNN and DEAGNN framework reported in the paper: Dista

172 Nov 08, 2022
An implementation of "Optimal Textures: Fast and Robust Texture Synthesis and Style Transfer through Optimal Transport"

Optex An implementation of Optimal Textures: Fast and Robust Texture Synthesis and Style Transfer through Optimal Transport for TU Delft CS4240. You c

Hans Brouwer 33 Jan 05, 2023
A solution to ensure Crowd Management with Contactless and Safe systems.

CovidTrack A Solution to ensure Crowd Management with Contactless and Safe systems. ML Model Mask Detection Social Distancing Detection Analytics Page

Om Khare 1 Nov 10, 2021
A denoising diffusion probabilistic model synthesises galaxies that are qualitatively and physically indistinguishable from the real thing.

Realistic galaxy simulation via score-based generative models Official code for 'Realistic galaxy simulation via score-based generative models'. We us

Michael Smith 32 Dec 20, 2022
Resilience from Diversity: Population-based approach to harden models against adversarial attacks

Resilience from Diversity: Population-based approach to harden models against adversarial attacks Requirements To install requirements: pip install -r

0 Nov 23, 2021
Implementation of BI-RADS-BERT & The Advantages of Section Tokenization.

BI-RADS BERT Implementation of BI-RADS-BERT & The Advantages of Section Tokenization. This implementation could be used on other radiology in house co

1 May 17, 2022
FusionNet: A deep fully residual convolutional neural network for image segmentation in connectomics

FusionNet_Pytorch FusionNet: A deep fully residual convolutional neural network for image segmentation in connectomics Requirements Pytorch 0.1.11 Pyt

Choi Gunho 102 Dec 13, 2022
POT : Python Optimal Transport

POT: Python Optimal Transport This open source Python library provide several solvers for optimization problems related to Optimal Transport for signa

Python Optimal Transport 1.7k Dec 31, 2022
A PyTorch implementation for Unsupervised Domain Adaptation by Backpropagation(DANN), support Office-31 and Office-Home dataset

DANN A PyTorch implementation for Unsupervised Domain Adaptation by Backpropagation Prerequisites Linux or OSX NVIDIA GPU + CUDA (may CuDNN) and corre

8 Apr 16, 2022
Integrated physics-based and ligand-based modeling.

ComBind ComBind integrates data-driven modeling and physics-based docking for improved binding pose prediction and binding affinity prediction. Given

Dror Lab 44 Oct 26, 2022
Code repository for our paper "Learning to Generate Scene Graph from Natural Language Supervision" in ICCV 2021

Scene Graph Generation from Natural Language Supervision This repository includes the Pytorch code for our paper "Learning to Generate Scene Graph fro

Yiwu Zhong 64 Dec 24, 2022
Align and Prompt: Video-and-Language Pre-training with Entity Prompts

ALPRO Align and Prompt: Video-and-Language Pre-training with Entity Prompts [Paper] Dongxu Li, Junnan Li, Hongdong Li, Juan Carlos Niebles, Steven C.H

Salesforce 127 Dec 21, 2022
InsightFace: 2D and 3D Face Analysis Project on MXNet and PyTorch

InsightFace: 2D and 3D Face Analysis Project on MXNet and PyTorch

Deep Insight 13.2k Jan 06, 2023
Official implementation of the ICCV 2021 paper: "The Power of Points for Modeling Humans in Clothing".

The Power of Points for Modeling Humans in Clothing (ICCV 2021) This repository contains the official PyTorch implementation of the ICCV 2021 paper: T

Qianli Ma 158 Nov 24, 2022
Official code for paper "Demystifying Local Vision Transformer: Sparse Connectivity, Weight Sharing, and Dynamic Weight"

Demysitifing Local Vision Transformer, arxiv This is the official PyTorch implementation of our paper. We simply replace local self attention by (dyna

138 Dec 28, 2022
利用yolov5和TensorRT从0到1实现目标检测的模型训练到模型部署全过程

写在前面 利用TensorRT加速推理速度是以时间换取精度的做法,意味着在推理速度上升的同时将会有精度的下降,不过不用太担心,精度下降微乎其微。此外,要有NVIDIA显卡,经测试,CUDA10.2可以支持20系列显卡及以下,30系列显卡需要CUDA11.x的支持,并且目前有bug。 默认你已经完成了

Helium 6 Jul 28, 2022
Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (CVPR 2021)

Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (CVPR 2021, official Pytorch implementatio

Microsoft 247 Dec 25, 2022