Karen Zhou | Fire Detection Using Keras

PythonC/C++BigQueryKeras

For my AP Research class, I tried to answer the following: "How can machine learning best be utilized to accurately classify video footage of forest fires on the average laptop computer?" In order to do so, I developed and implemented a convolutional neural network (CNN) in Google Colab using the Keras library. I also made use of methods in other libraries such as TensorFlow, NumPy, OpenCV, and matplotlib. In the collection of images below, I have included figures from my paper.

The CNN employed a binary classification method in order to identify images from either a fire or non-fire class, and I ended up training four different models in order to test optimal performance. In total, the final model trained on 3,760 images taken from Kaggle and other open source datasets. The highest F1 scores achieved was 0.96 for both the fire and non-fire classes.

Code ← Back to Portfolio

Gallery

class FireDetectionNet:
    @staticmethod

    def build(width, height, depth, classes):
        model = Sequential()
        inputShape = (height, width, depth)
        chanDim = -1

        model.add(SeparableConv2D(16, (7, 7), padding="same",
            input_shape=inputShape))
        model.add(Activation("relu"))
        model.add(BatchNormalization(axis=chanDim))
        model.add(MaxPooling2D(pool_size=(2, 2)))

        model.add(SeparableConv2D(32, (3, 3), padding="same"))
        model.add(Activation("relu"))
        model.add(BatchNormalization(axis=chanDim))
        model.add(SeparableConv2D(32, (3, 3), padding="same"))
        model.add(Activation("relu"))
        model.add(BatchNormalization(axis=chanDim))
        model.add(MaxPooling2D(pool_size=(2, 2)))

        model.add(Flatten())
        model.add(Dense(128))
        model.add(Activation("relu"))
        model.add(BatchNormalization())
        model.add(Dropout(0.5))

        model.add(Dense(classes))
        model.add(Activation("softmax"))

        return model

CNN layer implementation

import cv2
import numpy as np
from keras.models import load_model

def load_video(file_name):
    cam = cv2.VideoCapture("/content/" + file_name)
    currentframe = 0
    data = []

    while True:
        ret, frame = cam.read()
        if ret:
            image = cv2.resize(frame, (128, 128))
            image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
            data.append(image)
        else:
            break

    data = np.array(data, dtype="float32")

    cam.release()
    cv2.destroyAllWindows()

    return data / 255

def make_predictions(data, model_file):
    BATCH_SIZE = int((data.size) / 10)
    model = load_model('/content/' + model_file)
    predictions = model.predict(data, batch_size=BATCH_SIZE)

    return predictions

Video analysis implementation