def visualize_maps_features_2d(model, img):
fig = plt.figure(figsize=(30, 30))
extractor = Extractor(model)
extractor.activate()
img = cv.cvtColor(np.float32(img), cv.COLOR_BGR2RGB)
img = t.Compose([
t.ToPILImage(),
t.Resize((128, 128)),
t.ToTensor(),
t.Normalize(0.5, 0.5)])(np.uint8(img)).unsqueeze(0)
featuremaps = [extractor.CNN_layers[0](img)]
for x in range(1, len(extractor.CNN_layers)):
featuremaps.append(extractor.CNN_layers[x](featuremaps[-1]))
for x in range(len(featuremaps)):
layers = featuremaps[x][0, :, :, :].detach()
for i, filter in enumerate(layers):
if i == 64:
break
ax = fig.add_subplot(8, 8, i + 1)
ax.imshow(filter)
ax.axis('off')
ax.set_title("feature map: "+str(x)+" filter: "+str(i))
return fig
def visualize_maps_filters_1d(model):
fig = plt.figure(figsize=(30, 30))
extractor = Extractor(model)
extractor.activate()
for index, filter in enumerate(extractor.CNN_weights[0]):
ax = fig.add_subplot(8, 8, index + 1)
ax.plot(filter[0, :].cpu().detach())
ax.axis('off')
ax.set_title("filter: "+str(index), pad=10)
return fig
def visualize_maps_features_1d(model, img):
fig = plt.figure(figsize=(30, 30))
extractor = Extractor(model)
extractor.activate()
featuremaps = [extractor.CNN_layers[0](img)]
for x in range(1, len(extractor.CNN_layers)):
featuremaps.append(extractor.CNN_layers[x](featuremaps[-1]))
for x in range(len(featuremaps)):
layers = featuremaps[x][0, :, :].detach()
for i, filter in enumerate(layers):
if i == 64:
break
index = 130 +i+2
ax = fig.add_subplot(index)
ax.plot(filter)
ax.axis('off')
ax.set_title("feature map: "+str(x)+" filter: "+str(i))
return fig
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# print(model.children[0])
# Loading the model
# resnet = models.resnet50()
# extractor = Extractor(list(resnet.children()))
# model_children = list(model.children())
# print(type(model.children()))
# print(len(list(model.children())))
extractor = Extractor(list(model.modules()))
# print(model.modules())
extractor.activate()
# visualize the first conv layer filters
# plt.figure(figsize=(20, 17))
# for i, filter in enumerate(extractor.CNN_weights[0]):
# plt.subplot(8, 8, i+1) # (8, 8) because in conv0 we have 7x7 filters and total of 64 (see printed shapes)
# plt.imshow(filter[0, :, :].detach(), cmap='gray')
# plt.axis('off')
# plt.savefig('/home/ainedineen/blurry_vision/pytorch_untrained_models/imagenet/21x21FILTER_TEXT_resnet50_conv1_21.png')
# plt.show()
# # Visualising the filters
plt.figure(figsize=(35, 35))
for index, filter in enumerate(extractor.CNN_weights[0]):
plt.subplot(8, 8, index + 1)