Comments (9)
Hi zhongyuan, this is a very good question!
Inspired by the code from DualRCNet, we didn't use sub-pixel predictions for db-db pairs by using round()
. Therefore the "keypoints" will be strictly repeatable across images. I'll show you the reconstruction of Aachen in the image below. And the errors are reduced by the triangulation process and a high image resolution. We use sub-pixel predictions for db-query pairs, which contributes to the high localization accuracy.
from loftr.
Thanks for quick reply. It seems like continuous keyframes can still produce similar keypoints distribution due to round() function.
Following the localization strategy, the standard flow is: image retrieval -> 2D-2D match -> 3D-2D pnp -> pose result. Owing to round() during structure from motion, 3D points are binding with integer keypoints in keyframes. So during localization, how to handle the correspondence between 3D points and the new sub-pixel db keypoints or maybe the sup-pixel strategy is only used in query images ?
from loftr.
Correct. For LoFTR, we use sub-pixel strategy only for query images currently.
from loftr.
Thanks a lot. I'll have a try!
from loftr.
No problem 👏
from loftr.
Hi @Zhongyuan-Li @zehongs ,
I have a question about this part
"we use sub-pixel strategy only for query images currently."
Could you explain this part in more detail?
from loftr.
Hi,
For visual localization, one needs to first build a map with database images, and then establish correspondences for database-query images. Specifically, to find correspondences of the database points and query points.
Here, we round the point location when building database map. And we use sub-pixel strategy when computing the database-query crrespondences.
from loftr.
hi,@zehongs,i have a problem when I try to reconstruction my custom data by hloc+loftr. I used round() to save the result of loftr and inserted the pixel point as keypoint. Also I inserted the matching correspondence of two images. However, I got an error in pycolmap as check failed: bundle_adjuster.Solver(&reconstruction). Can you help to solve it if you have any experience? Many thanks
from loftr.
Hi, I have not encountered this problem. Maybe you can reduce the amount of data, and manually check if the output files are all correct for COLMAP.
from loftr.
Related Issues (20)
- training of smaller model
- Tensorboard visualization - are the indices correct?
- About train with fp16
- About the dataset indices of Megadepth HOT 2
- How to train the model for HPatches
- train model for image matching
- training on custom dataset HOT 4
- Can the model be limited to detecting key points in a region of interest?
- h5py can't read data
- meet an error when training loftr: EOFError: Ran out of input HOT 3
- image size
- Bug found in supervision process
- main_cfg problem
- How to train LoFTR on custom dataset?
- megadepth_indices下载不下来 HOT 1
- How to get image feature maps from the pretrained model?
- My dataset is only RGB, can I use it for training?
- ScanNet Training Pairs
- File "h5py\h5f.pyx", line 96, in h5py.h5f.open OSError: Unable to open file
- COLMAP dumps for Megadepth 1500 dataset
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from loftr.