Examples with Image ai.djl.modality.cv.Image used on opensource projects

Example 1 with Image

use of ai.djl.modality.cv.Image in project djl-demo by deepjavalibrary.

the class ImageClassification method main.

public static void main(String[] args) throws ModelException, IOException, TranslateException {
    // java default ImageIO doesn't work with GraalVM
    ImageFactory.setImageFactory(new GraalvmImageFactory());
    Image image;
    if (args.length == 0) {
        image = ImageFactory.getInstance().fromUrl(IMAGE_URL);
    } else {
        image = ImageFactory.getInstance().fromUrl(args[0]);
    }
    Criteria<Image, Classifications> criteria;
    if ("TensorFlow".equals(Engine.getInstance().getEngineName())) {
        Translator<Image, Classifications> translator = ImageClassificationTranslator.builder().addTransform(new Resize(224)).addTransform(new Normalize(MEAN, STD)).build();
        criteria = Criteria.builder().setTypes(Image.class, Classifications.class).optArtifactId("resnet").optTranslator(translator).optProgress(new ProgressBar()).build();
    } else {
        criteria = Criteria.builder().setTypes(Image.class, Classifications.class).optArtifactId("resnet").optProgress(new ProgressBar()).build();
    }
    try (ZooModel<Image, Classifications> model = ModelZoo.loadModel(criteria);
        Predictor<Image, Classifications> predictor = model.newPredictor()) {
        Classifications result = predictor.predict(image);
        System.out.println(result.toString());
    }
}

Example 2 with Image

use of ai.djl.modality.cv.Image in project djl-demo by deepjavalibrary.

the class WebCam method main.

public static void main(String[] args) throws IOException, ModelException, TranslateException {
    ZooModel<Image, DetectedObjects> model = loadModel();
    Predictor<Image, DetectedObjects> predictor = model.newPredictor();
    OpenCV.loadShared();
    VideoCapture capture = new VideoCapture(0);
    if (!capture.isOpened()) {
        System.out.println("No camera detected");
        return;
    }
    Mat image = new Mat();
    boolean captured = false;
    for (int i = 0; i < 10; ++i) {
        captured = capture.read(image);
        if (captured) {
            break;
        }
        try {
            Thread.sleep(50);
        } catch (InterruptedException ignore) {
        // ignore
        }
    }
    if (!captured) {
        JOptionPane.showConfirmDialog(null, "Failed to capture image from WebCam.");
    }
    ViewerFrame frame = new ViewerFrame(image.width(), image.height());
    ImageFactory factory = ImageFactory.getInstance();
    while (capture.isOpened()) {
        if (!capture.read(image)) {
            break;
        }
        Image img = factory.fromImage(image);
        DetectedObjects detections = predictor.predict(img);
        img.drawBoundingBoxes(detections);
        frame.showImage(toBufferedImage((Mat) img.getWrappedImage()));
    }
    capture.release();
    predictor.close();
    model.close();
    System.exit(0);
}

Example 3 with Image

use of ai.djl.modality.cv.Image in project djl-demo by deepjavalibrary.

the class CanaryTest method main.

public static void main(String[] args) throws IOException, ModelException, TranslateException {
    logger.info("");
    logger.info("----------Environment Variables----------");
    System.getenv().forEach((k, v) -> logger.info(k + ": " + v));
    logger.info("");
    logger.info("----------Default Engine----------");
    Engine.debugEnvironment();
    logger.info("");
    logger.info("----------Device information----------");
    int gpuCount = CudaUtils.getGpuCount();
    logger.info("GPU Count: {}", gpuCount);
    if (gpuCount > 0) {
        logger.info("CUDA: {}", CudaUtils.getCudaVersionString());
        logger.info("ARCH: {}", CudaUtils.getComputeCapability(0));
    }
    String djlEngine = System.getenv("DJL_ENGINE");
    if (djlEngine == null) {
        djlEngine = "mxnet-native-auto";
    }
    Device device = NDManager.newBaseManager().getDevice();
    if (djlEngine.contains("-native-cu") && !device.isGpu()) {
        throw new AssertionError("Expecting load engine on GPU.");
    } else if (djlEngine.startsWith("tensorrt")) {
        testTensorrt();
        return;
    } else if (djlEngine.startsWith("onnxruntime")) {
        testOnnxRuntime();
        return;
    } else if (djlEngine.startsWith("xgboost")) {
        testXgboost();
        return;
    } else if (djlEngine.startsWith("tflite")) {
        testTflite();
        return;
    } else if (djlEngine.startsWith("python")) {
        testPython();
        return;
    } else if (djlEngine.startsWith("dlr")) {
        testDlr();
        // similar to DLR, fastText and SentencePiece only support Mac and Ubuntu 16.04+
        testFastText();
        testSentencePiece();
        return;
    } else if (djlEngine.startsWith("paddle")) {
        testPaddle();
        return;
    }
    logger.info("");
    logger.info("----------Test inference----------");
    String url = "https://resources.djl.ai/images/dog_bike_car.jpg";
    Image img = ImageFactory.getInstance().fromUrl(url);
    String backbone = "resnet50";
    Map<String, String> options = null;
    if ("TensorFlow".equals(Engine.getInstance().getEngineName())) {
        backbone = "mobilenet_v2";
        options = new ConcurrentHashMap<>();
        options.put("Tags", "");
    }
    Criteria<Image, DetectedObjects> criteria = Criteria.builder().optApplication(Application.CV.OBJECT_DETECTION).setTypes(Image.class, DetectedObjects.class).optFilter("backbone", backbone).optOptions(options).build();
    try (ZooModel<Image, DetectedObjects> model = ModelZoo.loadModel(criteria)) {
        try (Predictor<Image, DetectedObjects> predictor = model.newPredictor()) {
            DetectedObjects detection = predictor.predict(img);
            logger.info("{}", detection);
        }
    }
}

Example 4 with Image

use of ai.djl.modality.cv.Image in project djl-demo by deepjavalibrary.

the class CanaryTest method testDlr.

private static void testDlr() throws ModelException, IOException, TranslateException {
    String os;
    if (System.getProperty("os.name").toLowerCase().startsWith("mac")) {
        os = "osx";
    } else if (System.getProperty("os.name").toLowerCase().startsWith("linux")) {
        os = "linux";
    } else {
        throw new AssertionError("DLR only work on mac and Linux.");
    }
    ImageClassificationTranslator translator = ImageClassificationTranslator.builder().addTransform(new Resize(224, 224)).addTransform(new ToTensor()).build();
    Criteria<Image, Classifications> criteria = Criteria.builder().setTypes(Image.class, Classifications.class).optApplication(Application.CV.IMAGE_CLASSIFICATION).optFilter("layers", "50").optFilter("os", os).optTranslator(translator).optEngine("DLR").optProgress(new ProgressBar()).build();
    String url = "https://resources.djl.ai/images/kitten.jpg";
    Image image = ImageFactory.getInstance().fromUrl(url);
    try (ZooModel<Image, Classifications> model = ModelZoo.loadModel(criteria);
        Predictor<Image, Classifications> predictor = model.newPredictor()) {
        Classifications classifications = predictor.predict(image);
        logger.info("{}", classifications);
    }
}

Example 5 with Image

use of ai.djl.modality.cv.Image in project djl-demo by deepjavalibrary.

the class CanaryTest method testTflite.

private static void testTflite() throws ModelException, IOException, TranslateException {
    if (System.getProperty("os.name").startsWith("Win")) {
        throw new AssertionError("TFLite only work on macOS and Linux.");
    }
    Criteria<Image, Classifications> criteria = Criteria.builder().setTypes(Image.class, Classifications.class).optEngine("TFLite").optFilter("dataset", "aiyDish").build();
    try (ZooModel<Image, Classifications> model = ModelZoo.loadModel(criteria);
        Predictor<Image, Classifications> predictor = model.newPredictor()) {
        Image image = ImageFactory.getInstance().fromUrl("https://resources.djl.ai/images/sachertorte.jpg");
        Classifications prediction = predictor.predict(image);
        logger.info(prediction.toString());
        if (!"Sachertorte".equals(prediction.best().getClassName())) {
            throw new AssertionError("Wrong prediction result");
        }
    }
}