I ran through all the colab steps and it worked pretty well the sample datasets but not good with custom data provided as follows:
with WBC:
with OASIS:
with my datasets(publicaly available):
I have my preprocessing codes as follows to be applied on the support images and support labels:
def process_img(path: None):
img = cv2.imread(path,cv2.IMREAD_GRAYSCALE)
img = cv2.resize(img,(128,128)).reshape(1, 128, 128)
img = img.astype(np.float32)
return img
def process_seg(path: None):
seg = cv2.imread(path,cv2.IMREAD_GRAYSCALE)
seg = cv2.resize(seg,(128,128)).reshape(1, 128, 128)
seg = seg.astype(np.float32)
return seg
processed_support_images = [T(process_img(image)) for image in _support_images]
processed_support_labels = [T(process_seg(label))for label in _support_labels]
support_images = torch.stack(processed_support_images).to(device)
support_labels = torch.stack(processed_support_labels).to(device)
I ran through all the colab steps and it worked pretty well the sample datasets but not good with custom data provided as follows: with WBC:
with OASIS:
with my datasets(publicaly available):
![tufts](https://github.com/JJGO/UniverSeg/assets/73964224/5ad5f6a9-ba8d-463d-bb20-f2441f13a767)
Is there any issue with my preprocessing code?