You can deploy this model and start earning money today!

mmirudoddi-123/
unet-image-segmentation

Name: unet-image-segmentation
Availability: InStock

mit

Views

No views yet

import os import shutil

Define the target directories

image_dir = "dataset/images" mask_dir = "dataset/masks"

Create the directories if they don't exist

os.makedirs(image_dir, exist_ok=True) os.makedirs(mask_dir, exist_ok=True)

Move the uploaded image to the image directory

Assuming 'img1.png.jpeg' was the last uploaded image and is in the current directory

if os.path.exists('img1.png.jpeg'): shutil.move('img1.png.jpeg', os.path.join(image_dir, 'img1.png.jpeg')) print(f"Moved img1.png.jpeg to {image_dir}/") else: print("img1.png.jpeg not found in current directory.")

Move the dummy mask to the mask directory

Now, re-run the cell where you instantiate the dataset (cell q-SoLWWz65Ij) and the training loop.

!pip install torch torchvision segmentation-models-pytorch import torch import segmentation_models_pytorch as smp import segmentation_models_pytorch as smp

model = smp.Unet( encoder_name="resnet34", encoder_weights="imagenet", classes=10, activation=None )

print(model) import torch print(torch.cuda.get_device_name(0)) loss_fn = torch.nn.CrossEntropyLoss() optimizer = torch.optim.Adam(model.parameters(), lr=0.001) batch_size = 4 height = 256 width = 256

Create dummy input images and masks

images = torch.randn(batch_size, 3, height, width) masks = torch.randint(0, 10, (batch_size, height, width)).long()

for epoch in range(5):

predictions = model(images)

loss = loss_fn(predictions, masks)

optimizer.zero_grad()
loss.backward()
optimizer.step()

print("Epoch:", epoch, "Loss:", loss.item())

import torch import numpy as np import cv2 import segmentation_models_pytorch as smp model = smp.Unet( encoder_name="resnet34", encoder_weights="imagenet", classes=10, activation=None )

print("Model created successfully") images = torch.randn(1,3,256,256) masks = torch.randint(0,10,(1,256,256)) loss_fn = torch.nn.CrossEntropyLoss() optimizer = torch.optim.Adam(model.parameters(), lr=0.001) for epoch in range(20):

predictions = model(images)

loss = loss_fn(predictions, masks)

optimizer.zero_grad()
loss.backward()
optimizer.step()

print("Epoch:", epoch, "Loss:", loss.item())

from torch.utils.data import Dataset

class SegmentationDataset(Dataset): def init(self, images, masks): self.images = images self.masks = masks

def __getitem__(self, idx):
    image = cv2.imread(self.images[idx])
    mask = cv2.imread(self.masks[idx], 0)

    image = image / 255.0
    return image, mask

def __len__(self):
    return len(self.images)

from google.colab import files

uploaded = files.upload()

for fn in uploaded.keys(): print(f'User uploaded file "{fn}" with length {len(uploaded[fn])} bytes') [13/03, 14:49] Lilyput: from google.colab import files

uploaded = files.upload()

for fn in uploaded.keys(): print(f'User uploaded file "{fn}" with length {len(uploaded[fn])} bytes') import os import cv2 import numpy as np

assuming image_paths is a list of image file paths( e.g., from uploadded files)

add mask_paths is a corresponding list of mask file paths

#example:using the previously uploaded image image_paths =["img1.png.jpeg"]

#create a dummy mask file for demonstration if it doesn't exist dummy_mask_path = "dummy_mask.png" if not os.path.exists(dummy_mask_path): dummy_mask_data = np.zeros((256, 256), dtype=np.uint8) cv2.imwrite(dummy_mask_path, dummy_mask_data) print("Dummy mask file created:", dummy_mask_path)

mask_paths = [dummy_mask_path] try: dataset=SegmentationDataset(image_paths,mask_paths) print("Dataset created successfully") print(f"number of item in dataser:{len(dataset)}")

if len(dataset)>0: first_image, first_mask = dataset[0] print("First image shape:", first_image.shape) print("First mask shape:", first_mask.shape) except Exception as e: print(f"Error creating dataset: {e}") import os import cv2 import torch from torch.utils.data import Dataset

class SegmentationDataset(Dataset):

def __init__(self, image_dir, mask_dir):
    self.image_dir = image_dir
    self.mask_dir = mask_dir
    self.images = os.listdir(image_dir)

def __len__(self):
    return len(self.images)

def __getitem__(self, index):

    img_path = os.path.join(self.image_dir, self.images[index])
    mask_path = os.path.join(self.mask_dir, self.images[index].replace(".jpg",".png"))

    image = cv2.imread(img_path)
    image = cv2.resize(image,(256,256))
    image = torch.tensor(image).permute(2,0,1).float()

    mask = cv2.imread(mask_path,0)
    mask = cv2.resize(mask,(256,256))
    mask = torch.tensor(mask).long()

    return image, mask

dataset = SegmentationDataset( image_dir="dataset/images", mask_dir="dataset/masks" ) from torch.utils.data import DataLoader

train_loader = DataLoader(dataset, batch_size=1, shuffle=True) import os import shutil

Define the target directories

image_dir = "dataset/images" mask_dir = "dataset/masks"

Create the directories if they don't exist

os.makedirs(image_dir, exist_ok=True) os.makedirs(mask_dir, exist_ok=True)

Move the uploaded image to the image directory

Assuming 'img1.png.jpeg' was the last uploaded image and is in the current directory

Move the dummy mask to the mask directory

if os.path.exists('dummy_mask.png'): shutil.move('dummy_mask.png', os.path.join(mask_dir, 'dummy_mask.png')) print(f"Moved dummy_mask.png to {mask_dir}/") else: print("dummy_mask.png not found in current directory.") # Now, re-run the cell where you instantinate the dataset (cell q-SoLWWz65Ij) and the training import segmentation_models_pytorch as smp

model = smp.Unet( encoder_name="resnet34", encoder_weights="imagenet", classes=10, activation=None ) model.eval()

Define a dummy test_image for demonstration

Example: A single random image with 3 channels and 256x256 dimensions

test_image = torch.randn(1, 3, 256, 256)

If you want to use a real image from the dataset, you could do:

if 'dataset' in globals() and len(dataset) >

test_image, _ = dataset[0]

test_image = test_image.unsqueeze(0) # Add batch dimension

pred = model(test_image)

pred = torch.argmax(pred, dim=1) import matplotlib.pyplot as plt

plt.imshow(pred[0]) plt.title("Predicted Segmentation") plt.show() import torch

Define a dummy mask for demonstration purposes

Ensure it has the same shape as pred and values within the class range (0-9 for 10 classes)

mask = torch.randint(0, 10, pred.shape).long()

intersection = (pred & mask).float().sum() union = (pred | mask).float().sum()

iou = intersection / union print("IoU:", iou) torch.save(model.state_dict(), "segmentation_model.pth")

mmirudoddi-123/unet-image-segmentation

Define the target directories

Create the directories if they don't exist

Move the uploaded image to the image directory

Assuming 'img1.png.jpeg' was the last uploaded image and is in the current directory

Move the dummy mask to the mask directory

Now, re-run the cell where you instantiate the dataset (cell q-SoLWWz65Ij) and the training loop.

Create dummy input images and masks

assuming image_paths is a list of image file paths( e.g., from uploadded files)

add mask_paths is a corresponding list of mask file paths

Define the target directories

Create the directories if they don't exist

Move the uploaded image to the image directory

Assuming 'img1.png.jpeg' was the last uploaded image and is in the current directory

Move the dummy mask to the mask directory

Define a dummy test_image for demonstration

Example: A single random image with 3 channels and 256x256 dimensions

If you want to use a real image from the dataset, you could do:

if 'dataset' in globals() and len(dataset) >

test_image, _ = dataset[0]

test_image = test_image.unsqueeze(0) # Add batch dimension

Define a dummy mask for demonstration purposes

Ensure it has the same shape as pred and values within the class range (0-9 for 10 classes)

mmirudoddi-123/
unet-image-segmentation