Spaces:
Sleeping
Sleeping
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([]) | |