Comments (7)
It's also worth noting that Wu's algorithm is good enough that we are in a regime where the best known algorithm probably won't gain us more than 2x performance, while parallel could gain us several times more than that.
from connected-components-3d.
The block based approach is undermined by multi-label. It depends on being able to summarize entire regions using the knowledge that all foreground pixels will match.
from connected-components-3d.
It might still be fun to implement block-based for binary images, which are an important class of operation.
from connected-components-3d.
OpenCV implements Grana et al's Optimized Block Based Decision Tree for binary images (apparently by that team themselves) and I tested it out on a medical YACCLAB dataset. Their algorithm was almost twice as fast, getting 470-493 MVx/sec vs about 250 MVx/sec from mine. Pretty impressive. I had assumed that the ceiling for a good algorithm was more like 20% rather than nearly 100%.
from connected-components-3d.
Was able to provide a better algorithm for 6-connected which makes me happy. I think the real secret of BBDT is the ability to make 1/4 fewer writes in the first pass. Will have to try that out.
from connected-components-3d.
BBDT and approaches derived from it are not useful for multilabel, however I suspect that the pixel prediction approach is readily compatible. Create a forest of trees for each situation and use goto
to select the correct tree for the next voxel. This can allow the elimination of a lot of duplicate work. 4 and 6 connected are prime candidates for this. 6-connected especially is about 2-3x worse than a null pass, so there's lots of room for improvement.
This technique is described in some detail in this paper: https://prittt.github.io/pub_files/2016acivs.pdf
For multi-label, we would use knowledge of the previous differently labeled pixel to remove pieces from the full decision tree rather than positively select information that is already known.
from connected-components-3d.
It would be cool if I integrated some of the crazy binary algorithms from YACCLAB. That group is on a tear.
from connected-components-3d.
Related Issues (20)
- Applying Dust and largest_k dtype output option HOT 2
- dust sugnature HOT 1
- Massive memory Leak HOT 7
- 1D Array of 4 Elements Incorrect HOT 5
- Cannot find reference 'dust' in 'cc3d.py' HOT 2
- Question on comparing individual lesions between two masks based on the cc3d.statistics output. HOT 1
- Additional metrics support HOT 2
- Does cc3d also work with memmory-mapped numpy arrays and array-like data? HOT 14
- cc3d.statistics["bounding_boxes"] are wrong HOT 1
- largest_k fails for transposed arrays HOT 6
- About the lastest_k function HOT 4
- Statistics output HOT 7
- IndexError is generated when using cc3d.connected_components HOT 4
- Label Statistics HOT 7
- Add support to release linux aarch64 wheels HOT 2
- Add Remove Dust Function HOT 2
- cc3d.statistics['bounding_boxes'] is contains floats instead of int as slice positions HOT 2
- voxel_connectivity_graph and contacts can not be applied in 2D label HOT 5
- Build Equivalence Table Two Z Slices at a Time
- cc3d.dust fails HOT 3
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 connected-components-3d.