File size: 1,066 Bytes
c65aa62 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 |
import numpy as np
from pytorch_grad_cam import EigenCAM
from pytorch_grad_cam.utils.image import show_cam_on_image
import matplotlib.pyplot as plt
def generate_gradcam(model, target_layers, images, use_cuda=True, transparency=0.6):
results = []
targets = None
cam = EigenCAM(model, target_layers, use_cuda=use_cuda)
for image in images:
input_tensor = image.unsqueeze(0)
grayscale_cam = cam(input_tensor, targets=targets)
grayscale_cam = grayscale_cam[0, :]
img = input_tensor.squeeze(0).to("cpu")
rgb_img = np.transpose(img, (1, 2, 0))
rgb_img = rgb_img.numpy()
cam_image = show_cam_on_image(
rgb_img, grayscale_cam, use_rgb=True, image_weight=transparency
)
results.append(cam_image)
return results
def visualize_gradcam(images, figsize=(10, 10), rows=2, cols=5):
fig = plt.figure(figsize=figsize)
for i in range(len(images)):
plt.subplot(rows, cols, i + 1)
plt.imshow(images[i])
plt.xticks([])
plt.yticks([])
|