cnn_fire

Convolutional neural network to detect fire

Scripts

• cnn_architectures.py - Defines several CNN architectures;
• fcmmanager.py - Sends push notifications to firebase service
• googlemanager.py - Manages the connection with goodle drive service
• nn_test.py - Class to handle testing one image or video in one previously recorded CNN train
• raw_images_loader.py - Load raw images into numpy arrays to be consumed by CNN
• test_on_cam.py - Apply treined CNN on image captured by camera
• train_manual.py - Train the network manually. Read from raw images, augmenting manualy etc
• train.py - Train network with Keras generator
• confusion_matrix.py - Plot the confusion matrix for a given trained network
• plot_filtesrs.py - Plot treined network learned filters
• resizeImagesInFolder.py - Resize images in one folder
• renameimagesindir.py - Rename the images in one dir so that we can distinguish between them on training time
• images_downloader.py - Download images from some services on internet
• deleteDuplicateImages.py - Deletes duplicate images in a given folder

Files

• model.json - The stores json architecture from training with its parameters setted.
• wights.h5 - The network treined weights

Packages Reqs:

• Keras
• TensorFlow
• Numpy
• Matplotlib
• Scipy
• ApiClient
• Imutils
• OpenCv
• Argparse
• MoviePy
• SimpleJson
• PyFcm
• Oauth2client
• Httplib2