search

AIBluefisher / DAGSfM

Distributed and Graph-based Structure from Motion. This project includes the official implementation of our Pattern Recognition 2020 paper: Graph-Based Parallel Large Scale Structure from Motion.
https://aibluefisher.github.io/GraphSfM/
BSD 3-Clause "New" or "Revised" License
395 stars 84 forks source link

Large data set processing #28

Closed Yzhbuaa closed 3 years ago

Yzhbuaa commented 3 years ago

Hello, I wonder how many images can GraphSfM process at a time. Does GraphSfM have the potential to process millions of images at a time? Thank you.

AIBluefisher commented 3 years ago

It's no doubt that GraphSfM can handle millions of images, as long as you partition images to clusters that your computer can handle. But it also has requirements of your computer. The main limitation is the great memory requirement for image clustering, as I use NCut or spectral clustering first to solve the graph cut problem. You can see from the unit test in clustering module, I test at most 10,000 images with 1,000,000 edges on my computer with 16GB RAM. For more larger scale datasets, we can try other stategies to solve the graph cut problem. But you still need to ensure that your computer's memory is large enough.

Yzhbuaa commented 3 years ago

Thank you for your detailed explanation!