Fork me on GitHub

Bytedeco makes native libraries available to the Java platform by offering ready-to-use bindings generated with the codeveloped JavaCPP technology. This, we hope, is the missing bridge between Java and C/C++, bringing compute-intensive science, multimedia, computer vision, deep learning, etc to the Java platform.

Core Technologies

  • JavaCPP [API] – A tool that can not only generate JNI code but also build native wrapper library files from an appropriate interface file written entirely in Java. It can also parse automatically C/C++ header files to produce the required Java interface files.

Prebuilt Java Bindings to C/C++ Libraries

These are part of a project that we call the JavaCPP Presets. Many coexist in the same GitHub repository, and all use JavaCPP to wrap predefined C/C++ libraries from open-source land. The bindings expose almost all of the relevant APIs and make them available in a portable and user-friendly fashion to any Java virtual machine (including Android), as if they were like any other normal Java libraries. We have presets for the following C/C++ libraries:

We will add more to this list as they are made, including those from outside the bytedeco/javacpp-presets repository.

Projects Leveraging the Presets Bindings

  • JavaCV [API] – Library based on the JavaCPP Presets that depends on commonly used native libraries in the field of computer vision to facilitate the development of those applications on the Java platform. It provides easy-to-use interfaces to grab frames from cameras and audio/video streams, process them, and record them back on disk or send them over the network.
  • JavaCV Examples – Collection of examples originally written in C++ for the book entitled OpenCV 2 Computer Vision Application Programming Cookbook by Robert Laganière, but ported to JavaCV and written in Scala.
  • ProCamCalib – Sample JavaCV application that can perform geometric and photometric calibration of a set of video projectors and color cameras.
  • ProCamTracker – Another sample JavaCV application that uses the calibration from ProCamCalib to implement a vision method that tracks a textured planar surface and realizes markerless interactive augmented reality with projection mapping.

More Project Information

Please refer to the contribute and download pages for more information about how to help out or obtain this software.

See the developer site on GitHub for more general information about the Bytedeco projects.

Latest News feed twitter

Bytedeco as a distribution

July 17, 2018

With version 1.4.2 now released, it’s time to publish a small blog post introducing the main innovations. As everyone probably knows, we’ve also released version 1.4.1 back in March. Thanks to the CI servers at AppVeyor and Travis CI, we are now able to produce full releases of all binaries without too much effort, so we will probably be maintaining a release cycle of about 3 months. This brings Bytedeco one step closer to a distribution such as Anaconda, Fedora, or Ubuntu, but cross-platform and for multiple languages (that is being able to access native libraries from other languages than C++), where the presets for OpenCV and TensorFlow, for example, now also bundle the official Java APIs, but still provide users with access to the C/C++ APIs as well as additional features for better integration. PointerScope is one such new feature that can help manage native memory, even when using multiple different APIs together.

Until now, to prevent garbage from hanging around too long in memory, we had to call manually Pointer.deallocate() on all native objects (either directly or with try-with-resources statements). We can now use a PointerScope to manage a group of such objects, which is similar to workspaces in Deeplearning4j (but without the workspace memory) or Scope from Panama, and as exemplified below for OpenCV and TensorFlow, a single PointerScope works even across libraries:

    static void checkStatus(Status s) {
        if (!s.ok()) {
            throw new RuntimeException(s.error_message().getString());

    static void detectObjects(File file, Session session) throws Exception {
        // load the image file
        Mat mat = imread(file.getCanonicalPath());
        if (mat == null || mat.empty()) return;
        cvtColor(mat, mat, CV_BGR2RGB);
        Tensor tensor = new Tensor(DT_UINT8,
            new TensorShape(1, mat.rows(), mat.cols(), mat.channels()),

        // run inference
        String[] inputNames = {"image_tensor"};
        Tensor[] inputTensors = {tensor};
        String[] outputNames = {"detection_boxes", "detection_scores", "detection_classes", "num_detections"};
        TensorVector outputTensors = new TensorVector();
        checkStatus(session.Run(new StringTensorPairVector(inputNames, inputTensors),
                new StringVector(outputNames), new StringVector(), outputTensors));

        // do something with outputTensors...

    public static void main(String args[]) throws Exception {
        // read the model
        GraphDef graph = new GraphDef();
        checkStatus(ReadBinaryProto(Env.Default(), "/path/to/frozen_inference_graph.pb", graph));

        // create the session
        Session session = new Session(new SessionOptions());

        // detect objects from all files in directory
        File dir = new File("/path/to/images/");
        for (File f : dir.listFiles()) {
            try (PointerScope scope = new PointerScope()) {
                detectObjects(f, session);

When executing this code with opencv-platform, tensorflow-platform, and the model file from ssd_inception_v2_coco_2017_11_17.tar.gz, the output shows that memory usage rapidly stabilizes around 1 GB:


However, if we take the PointerScope out of the loop, it goes on to fill up all the memory that we give it, which can easily exceed 16 GB by default on today’s machines:


This is good not only to prevent wasting resources, but also to limit memory fragmentation. If you are interested in using TensorFlow this way, thanks to Nico Hezel, who also fixed the build for TensorFlow with CUDA support on Windows, additional samples are now available at:

Finally, other important changes include:

Please find more information about these updates and other changes in the files for JavaCPP, JavaCPP Presets, JavaCV, ProCamCalib, and ProCamTracker. Many thanks to all contributors! According to git shortlog 1.4..1.4.2 --summary they are:

agiasmothi, Alex Black, Aman Gupta, canelzio, delthas, gunther82, Jeremy Apthorp, Jeremy Laviole, Johan Vos, Kevin Watters, louxiu, Mariano Scazzariello, Michael Haslgrübler, mifritscher, Nico Hezel, SIY1121, Sören Brunk, Stanislav Chizhik, Vincent Baines (vb216)

For completeness, here are all contributors since the beginning according to git shortlog --summary:

Adam Gibson, agiasmothi, Agustin Alvarez, Alex Black, alicana, Aman Gupta, Amnon Owed, Amos Yuen, Andreas Eberle, Arseniy Pendryak, Ashley, beligum, Benjamin Hill, bennyveo, canelzio, Chetan Narsude, Chris Chow, Chris Nuernberger, Cyprien Noel (cypof), delthas, Djalma Antonio dos Passos Filho, Dmitriy Gerashenko, dragos-d, Edison Wang, Edu Garcia, egorapolonov, emilio1986, Fabrizio (Misto) Milo, Felix Andrews, Florian Enner, François Garillot, Fredrik Henricsson, Gabor Nagy, George Kankava, Gertjan Al, GroG (supertick), gunther82, HungHD, Jadon Fowler, Jarek Sacha (jarek), Jeremy Apthorp, Jeremy Laviole, Jeremy Martin, Jeroen Arens, jjYBdx4IL, Johan Vos, John Sichi, Jonathan Leitschuh, juergenuhl, kenji yoshida, Kevin Watters, leef, lfdversluis, Lloyd Chan (lloydmeta), louxiu, Luca Barbato, mabruce, Mariano Scazzariello, Martin, matrupihai, Maurice, Maurice Betzel (mbetzel), Michael Dietz, Michael Haslgrübler, Michael Tandy, Michal Conos, mifritscher, Miroslav Zoričák, mmanco, Mo Tao, mtandy, Nico Hezel, n-kai-cj, Oleg Kuliasov, Pachev Joseph, Paolo Bolettieri (paolobolettieri), Paul Gregoire, Paul Hammant, Paul Heideman, Pawel Hajduk, pchundi, Perry Nguyen, Philipp Moritz, Piotr, Richard Alam, Ryan Osial, Sam Carlberg, Samuel, Sean Carlisle, SIY1121, Sören Brunk, Stanislav Chizhik, Tatsuyuki Ishi, Teerapap Changwichukarn, The Gitter Badger, Tiago Daniel Jacobs, Tom Clark, Trejkaz (pen name), TwistedUmbrella, Vincent Baines (vb216), waldemarnt, Yizhi Liu, zhanhb, wmz7year, 李立强

This list does not include anyone who has helped in other ways, including but not limited to fixing CI settings, testing builds, debugging code, filing issues, making corrections, reporting bugs, proposing ideas, updating the wiki, etc, but please do continue to communicate with us via the mailing list from Google Groups, issues on GitHub, or the chat room at Gitter. Your input is valuable, we hope to continue hearing from all of you!