Repository files navigation

PopSift is an open-source implementation of the SIFT algorithm in CUDA.PopSift tries to stick as closely as possible to David Lowe's famous paper [1], while extracting features from an image in real-time at least on an NVidia GTX 980 Ti GPU.

Check out thedocumentation for more info.

PopSift compiles and works with NVidia cards of compute capability >= 3.0 (including the GT 650M), but the code is developed with the compute capability 5.2 card GTX 980 Ti in mind.

CUDA SDK 11 does no longer support compute capability 3.0. 3.5 is still supported with deprecation warning.

PopSift depends on:

• Host compiler that supports C++14 for CUDA SDK >= 9.0 and C++11 for CUDA SDK 8

• CUDA >= 8.0

Optionally, for the provided applications:

• Boost >= 1.71 (required components {atomic, chrono, date-time, system, thread}-dev)

• DevIL (libdevil-dev) can be used to load a broader range of image formats, otherwise only pgm is supported.

In order to build the library you can run:

mkdir build && cd buildcmake ..makemake install

Some build options are available:

• PopSift_BUILD_EXAMPLES (default:ON) enable building the applications that showcase the use of the library.

• BUILD_SHARED_LIBS (default:ON) controls the type of library to build (ON for shared libraries,OFF for static)

The main artifact created islibpopsift.If enabled, the test applicationpopsift-demo is created as well.Callingpopsift-demo without parameters shows the options.

To integrate PopSift into other software, link withlibpopsift.If your are using CMake for building your project you can easily add PopSift to your project.Once you have built and installed PopSift in a directory (say,<prefix>), in yourCMakeLists.txt file just add the dependency

# Find the package from the PopSiftConfig.cmake# in <prefix>/lib/cmake/PopSift/. Under the namespace PopSift::# it exposes the target popsift that allows you to compile# and link with the libraryfind_package(PopSift CONFIG REQUIRED)...# suppose you want to try it out in a executableadd_executable(poptest yourfile.cpp)# add link to the librarytarget_link_libraries(poptestPUBLIC PopSift::popsift)

Then, in order to build just pass the location ofPopSiftConfig.cmake from the cmake command line:

cmake .. -DPopSift_DIR=<prefix>/lib/cmake/PopSift/

The caller must create apopart::Config struct (documented insrc/sift/sift_conf.h) to control the behaviour of the PopSift, and instantiate an object of classPopSift (found insrc/sift/popsift.h).

After this, images can be enqueued for SIFT extraction using (enqueue()).A valid input is a single plane of grayscale values located in host memory.They can passed as a pointer to unsigned char, with a value range from 0 to 255, or as a pointer to float, with a value range from 0.0f to 1.0f.

Only host memory limits the number of images that can be enqueued.Theenqueue function returns a pointer to aSiftJob immediately and performs the feature extraction asynchronously.The memory of the image passed to enqueue remains the caller's responsibility. CallingSiftJob::get on the returned job blocks until features are extracted, and returns them.

Features offer iterators that iterate over objects of typeFeature.Both classes are documented insift_extremum.h.Each feature represents a feature point in the coordinate system of the input image, providing X and Y coordinates and scale (sigma), as well as several alternative descriptors for the feature point (according to Lowe, 15% of the feature points should be expected to have 2 or more descriptors).

In an alternate, deprecated, blocking API,init() must be called to pass image width and height to PopSift, followed by a call toexecuted() that takes image data and returns the extracted features.execute() is synchronous and blocking.

As far as we know, no implementation that is faster than PopSift at the time of PopSift's release comes under a license that allows commercial use and sticks close to the original paper at the same time as well.PopSift can be configured at runtime to use constants that affect it behaviours.In particular, users can choose to generate results very similar to VLFeat or results that are closer (but not as close) to the SIFT implementation of the OpenCV extras.We acknowledge that there is at least one SIFT implementation that is vastly faster, but it makes considerable sacrifices in terms of accuracy and compatibility.

• master branch on Linux.

• develop branch on Linux.

• develop branch on Windows.

PopSift is licensed underMPL v2 license.SIFT was patented in the United States from 1999-03-08 to 2020-03-28. See thepatent link for more information.PopSift license only concerns the PopSift source code and does not release users of this code from any requirements that may arise from patents.

If you use PopSift for your publication, please cite us as:

@inproceedings{Griwodz2018Popsift,author ={Griwodz, Carsten and Calvet, Lilian and Halvorsen, P{\aa}l},title ={Popsift: A Faithful SIFT Implementation for Real-time Applications},booktitle ={Proceedings of the 9th {ACM} Multimedia Systems Conference},series ={MMSys '18},year ={2018},isbn ={978-1-4503-5192-8},location ={Amsterdam, Netherlands},pages ={415--420},numpages ={6},doi ={10.1145/3204949.3208136},acmid ={3208136},publisher ={ACM},address ={New York, NY, USA},}

PopSift was developed within the projectPOPART, which has been funded by theEuropean Commission in the Horizon 2020 framework.

[1]: Lowe, D. G. (2004). Distinctive Image Features from Scale-Invariant Keypoints. International Journal of Computer Vision, 60(2), 91–110. doi:10.1023/B:VISI.0000029664.99615.94