Upsampling Map¶

Upsampling map is a way to visualize which regions of the image lead to the high unit activations.

Before generating the upsampling map, you need to ensure that the images can be highly activated by the given unit in DNN, which can be obtained using dnn_topstim (Select the topK stimuli from a stimulus set).

The procedures are as follows. First, we feed the image into the network and get its feaure map at the given channel. Then we use some interpolate method to upsample the feaure map into the original image space. After setting the threshold to filter the map, we finally get the upsampling map.

There is an example of up-sampling(us)method through using python library of DNNBrain.

The original image used in this doc is displayed as below:

original

Example¶

import numpy as np
import matplotlib.pyplot as plt

from dnnbrain.dnn.base import ip
from dnnbrain.dnn.models import AlexNet
from dnnbrain.dnn.algo import UpsamplingActivationMapping

# Prepare DNN and image
dnn = AlexNet()
image = plt.imread('ILSVRC_val_00095233.JPEG')

# Using up-sampling(us) method to display
# regions of the image that contribute to
# the activation of the 122th unit of conv5.
up_estimator =UpsamplingActivationMapping(dnn, 'conv5', 122)
up_estimator.set_params(interp_meth='bicubic', interp_threshold=0.95)
img_out = up_estimator.compute(image)

# transform to PIL image and save out
img_out = ip.to_pil(img_out, True)
img_out.save('ILSVRC_val_00095233_rf_us.JPEG')

The receptive field upsampling is displayed as below:

vanilla