I am trying to generate masks for the image segmentation task using dinov2 model. I have generated some using patch embeddings but when I try to save these mask images, the saved mask images are all black. How can i resolve this error?
import numpy as np
import torch.nn.functional as F
ans=[]
masks=[]
for idx,(image,_) in enumerate(trainloader):
seg=segmentor(1)
x=seg.forward(layerpatches)
print(f"mask image shape:{x.shape}")
for i in range(x.shape[0]):
image_array = x[i].detach().squeeze().numpy()
mask_image = image_array
image=image.permute(0,2,3,1).squeeze()
fig, ax = plt.subplots(1, 2, figsize=(10, 5))
ax[0].imshow(mask_image)
ax[0].set_title('Mask')
ax[0].axis('off')
ax[1].imshow(image)
ax[1].set_title('Corresponding Image')
ax[1].axis('off')
plt.show()
mask_name = f"{img_name[idx]}_mask.jpg"
mask_np = np.clip(mask_rgb, 0, 255).astype(np.uint8)
# # Save the adjusted mask as PNG
mask_path = os.path.join(mask_dir, mask_name)
cv2.imwrite(mask_path,mask_image)
break
The mask and corresponding image visualized on kaggle Same mask when saved in either png or jpg format
I have tried all the techniques to save the image as well as converting mask to RGB but saved images are all black.
