Unable to interpret feature maps

I have created network to classify emotions of the people using CAER dataset.
For the model I am using resnet architecture (not the pretrained one) and trying to save feature map for the last convolutional layer of last block.
I am able to save the maps but when I try to view them its quite difficult to interpret not sure whether this is expected or I am doing something wrong in either capturing or saving them.
Below is the code and feature maps.

Visualize code

writer = SummaryWriter('runs/caer')

class LayerActivations():

    def __init__(self,model,layer_num):
        self.hook = model.layer4[layer_num].conv2.register_forward_hook(self.hook_fn)

    def hook_fn(self,module,input,output):
        self.features = output.detach()

    def remove(self):

def imshow_image(image):

    mean = torch.tensor([0.370, 0.225, 0.189])
    std = torch.tensor([0.172, 0.130, 0.122])
    for img,m,s in zip(image,mean,std):


def load_checkpoint(filepath):
    checkpoint = torch.load(filepath)
    model = checkpoint['model']

return model

def show_activation(train_loader,dev):
    dataiter = iter(train_loader)
    images,labels = dataiter.next()


    model = load_checkpoint('caer_model.pth')
    features = model.module

    conv_out = LayerActivations(features,1)

    out_pred = model(images[0:2].to(dev))
    act = conv_out.features

    act = torch.nn.Upsample(size=(224,224),mode='bilinear')(act)
    for i in range(512):
        filename = './features/feature_'+str(i)+'.jpg'     
        img = act[0][i] 
        torchvision.utils.save_image(img, filename, nrow=8 )

Feature maps:-

Note:- I am able to post only one feature map.

I think Grad Cam will be useful for you. It is a visualization technique for understanding which areas are dominant for CNN model decesion. Here is an example from repo:

ok thanks.
But I wanted to know whether my code is correct or not?