-
Notifications
You must be signed in to change notification settings - Fork 25
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Changing neural network architecture doesn't improve heatmaps #3
Comments
model.py import tensorflow as tf class MNIST_CNN: def init(self, name='MNIST_CNN'): def convlayer(self, input, shape, name): def fclayer(self, input, shape, name, prop=True): def call(self, images, reuse=False):
@Property |
train.py program from utils import DataGenerator, MNISTLoader import tensorflow as tf logdir = './logs/' class Trainer:
if name == 'main': |
utils.py program import gzip DATA_PATH = './mnist_png/training' class DataGenerator:
class MNISTLoader:
if name == 'main':
|
lrp.py program from utils import MNISTLoader import numpy as np logdir = './logs/' class LayerwiseRelevancePropagation:
def get_relevances(self):
def backprop_fc(self, name, activation, relevance): def backprop_flatten(self, activation, relevance): def backprop_max_pool2d(self, activation, relevance, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1]): def backprop_conv2d(self, name, activation, relevance, strides=[1, 1, 1, 1]): def get_heatmap(self,i):
def test(self):
if name == 'main':
|
In mnist program, I modified it to feed images and changed neural network architecture by including more convolutional layers and tried for cat/Dog images instead of mnist data. I got heat maps which include features other than cat and dog also. please let me know what has to be done for getting proper heat maps.
The text was updated successfully, but these errors were encountered: