Super fast implementation of ICP in CUDA for compute capable devices 2.0 or higher. On an nVidia GeForce GTX TITAN X it runs at over 750Hz (using projective data assocation). To compile all architectures you'll need CUDA 7.0 I think, (or 6.5 with the special release for 9xx cards). You can compile for older cards by removing the unsupported architectures from the CMakeLists.txt file.
Requires CUDA, Pangolin, Eigen and Sophus. I've built it to take in raw TUM RGB-D datasets to do frame-to-frame dense ICP as an example application.
The particular version of ICP implemented is the one introduced by KinectFusion. This means a three level coarse-to-fine registration pyramid, from 160x120 to 320x240 and finally 640x480 image sizes, with 4, 5 and 10 iterations per level respectively.
Run like;
./ICP ~/Desktop/rgbd_dataset_freiburg1_desk/ -v
Where ~/Desktop/rgbd_dataset_freiburg1_desk/ contains the depth.txt file, for more information see here.
The main idea to getting the best performance is determining the best thread/block sizes to use. I have provided an exhaustive search function to do this, since it varies between GPUs. Simply pass the "-v" switch to the program to activate the search. The code will then first do a search for the best thread/block sizes and then run ICP and output something like this on an nVidia GeForce GTX TITAN X;
GeForce GTX TITAN X
Searching for the best thread/block configuration for your GPU...
Best: 256 threads, 96 blocks (1.3306ms), 100%
ICP: 1.3236ms
ICP speed: 755Hz
The code will output one file; output.poses. You can evaluate it on the TUM benchmark by using their tools. I get something like this;
python ~/stuff/Kinect_Logs/Freiburg/evaluate_ate.py ~/Desktop/rgbd_dataset_freiburg1_desk/groundtruth.txt output.poses
0.144041
The difference in values comes down to the fact that each method uses a different reduction scheme and floating point operations are not associative.
Also, if you're using this code in academic work and it would be suitable to do so, please consider referencing some of my possibly relevant research in your literature review/related work section.
icpcuda's People
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
๐ 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. ๐๐๐
-
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 โค๏ธ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.