Keras-RCNN

keras-rcnn is the Keras package for region-based convolutional neural networks.

Getting Started

Let’s read:

training, validation, test = keras_rcnn.datasets.malaria_phenotypes.load_data()

classes = {
    "rbc": 1, "leu": 2, "ring": 3, "tro": 4, "sch": 5, "gam": 6
}

generator = keras_rcnn.preprocessing.ObjectDetectionGenerator()

generator = generator.flow(training, classes)

validation_data = keras_rcnn.preprocessing.ObjectDetectionGenerator()

validation_data = validation_data.flow(validation, classes)

and inspect our training data:

(target_bounding_boxes, target_image, target_scores, _), _ = generator.next()

target_bounding_boxes = numpy.squeeze(target_bounding_boxes)

target_image = numpy.squeeze(target_image)

target_scores = numpy.argmax(target_scores, -1)

target_scores = numpy.squeeze(target_scores)

_, axis = matplotlib.pyplot.subplots(1, figsize=(12, 8))

axis.imshow(target_image)

for target_index, target_score in enumerate(target_scores):
    if target_score > 0:
        xy = [
            target_bounding_boxes[target_index][0],
            target_bounding_boxes[target_index][1]
        ]

        w = target_bounding_boxes[target_index][2] - target_bounding_boxes[target_index][0]
        h = target_bounding_boxes[target_index][3] - target_bounding_boxes[target_index][1]

        rectangle = matplotlib.patches.Rectangle(xy, w, h, edgecolor="r", facecolor="none")

        axis.add_patch(rectangle)

matplotlib.pyplot.show()

Let’s create an RCNN instance:

image = keras.layers.Input((None, None, 3))

model = keras_rcnn.models.RCNN(image, classes=len(classes) + 1)

and pass our preferred optimizer to the compile method:

optimizer = keras.optimizers.Adam(0.0001)

model.compile(optimizer)

Finally, let’s use the fit_generator method to train our network:

model.fit_generator(generator)

Slack

We’ve been meeting in the #keras-rcnn channel on the keras.io Slack server.

You can join the server by inviting yourself from the following website:

https://keras-slack-autojoin.herokuapp.com/