U-Net architecture in Computer Vision - Model Metrics & Evaluation

Metrics & Evaluation - U-Net architecture

Which metric matters for U-Net architecture and WHY

U-Net is mainly used for image segmentation. This means it labels each pixel as part of an object or background. The key metrics are Dice coefficient and Intersection over Union (IoU). They measure how well the predicted mask matches the true mask. Accuracy alone can be misleading because most pixels might be background. Dice and IoU focus on overlap, which is what matters for segmentation quality.

Confusion matrix for pixel-wise segmentation

      | Predicted Positive | Predicted Negative |
      |--------------------|--------------------|
      | True Positive (TP)  | False Negative (FN) |
      | False Positive (FP) | True Negative (TN)  |

    Total pixels = TP + FP + TN + FN

    Dice coefficient = 2 * TP / (2 * TP + FP + FN)
    IoU = TP / (TP + FP + FN)
    

This matrix counts pixels, not images. TP means pixels correctly labeled as object. FP means pixels wrongly labeled as object. FN means object pixels missed.

Precision vs Recall tradeoff in U-Net segmentation

Precision means how many predicted object pixels are correct. High precision means few false positives (wrongly labeled pixels).

Recall means how many true object pixels were found. High recall means few false negatives (missed pixels).

For medical images, missing a tumor pixel (low recall) is worse than marking some extra pixels (lower precision). So recall is more important.

For autonomous driving, marking too many pixels as obstacles (low precision) can cause unnecessary stops. So precision matters more.

Dice and IoU balance precision and recall by measuring overlap.

Good vs Bad metric values for U-Net segmentation

• Good: Dice > 0.8 and IoU > 0.7 usually mean the model segments objects well.
• Bad: Dice < 0.5 or IoU < 0.4 means poor overlap, many pixels are wrong.
• High accuracy (>90%) can be misleading if background dominates and object pixels are few.

Common pitfalls in evaluating U-Net segmentation

• Accuracy paradox: High accuracy but poor segmentation if background pixels dominate.
• Data leakage: Using test images similar to training can inflate metrics.
• Overfitting: Very high training Dice but low test Dice means model memorizes training masks.
• Ignoring class imbalance: Small objects may be missed if metrics do not focus on them.

Self-check: Your U-Net model has 98% accuracy but 12% recall on tumor pixels. Is it good?

No, it is not good. The high accuracy likely comes from many background pixels correctly labeled. But 12% recall means the model misses 88% of tumor pixels. This is dangerous in medical use because tumors are not detected. You need to improve recall even if accuracy drops.

Key Result

Dice coefficient and IoU are key metrics for U-Net segmentation because they measure pixel overlap, balancing precision and recall.

Practice

1. What is the main purpose of the U-Net architecture in computer vision?

easy

A. To compress images for storage

B. To classify images into categories

C. To find exact shapes and boundaries in images

D. To generate new images from noise

Solution

1. Step 1: Understand U-Net's design goal

   U-Net is designed to segment images, meaning it finds exact shapes and boundaries.

2. Step 2: Compare with other tasks

   Classification assigns labels, generation creates images, compression reduces size; these are not U-Net's main tasks.

3. Final Answer:

   To find exact shapes and boundaries in images -> Option C

4. Quick Check:

   U-Net = segmentation [OK]

Hint: U-Net = image segmentation for shapes [OK]

Common Mistakes:

• Confusing segmentation with classification
• Thinking U-Net generates images
• Assuming U-Net compresses images

2. Which of the following correctly describes the structure of U-Net?

easy

A. Two paths: one shrinks and one expands the image size

B. A single path that shrinks the image size

C. A path that only expands the image size

D. Multiple independent paths without connection

Solution

1. Step 1: Recall U-Net architecture parts

   U-Net has an encoder path that shrinks and a decoder path that expands the image size.

2. Step 2: Eliminate incorrect options

   Single shrinking path or only expanding path do not match U-Net. Independent paths without connection are incorrect.

3. Final Answer:

   Two paths: one shrinks and one expands the image size -> Option A

4. Quick Check:

   U-Net = encoder + decoder [OK]

Hint: U-Net has encoder and decoder paths [OK]

Common Mistakes:

• Thinking U-Net only shrinks images
• Believing U-Net only expands images
• Ignoring the connection between paths

3. Consider this simplified U-Net code snippet in Python using PyTorch:

import torch
import torch.nn as nn

class SimpleUNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.down = nn.Conv2d(1, 16, 3, padding=1)
        self.up = nn.ConvTranspose2d(16, 1, 2, stride=2)

    def forward(self, x):
        x1 = self.down(x)  # shrink
        x2 = self.up(x1)   # expand
        return x2

model = SimpleUNet()
input_tensor = torch.randn(1, 1, 64, 64)
output = model(input_tensor)
print(output.shape)

What will be the printed output shape?

medium

A. torch.Size([1, 1, 64, 64])

B. torch.Size([1, 1, 128, 128])

C. torch.Size([1, 16, 64, 64])

D. torch.Size([1, 16, 32, 32])

Solution

1. Step 1: Analyze down convolution output size

   Input is 64x64, Conv2d with kernel=3, padding=1 keeps size 64x64, output channels=16.

2. Step 2: Analyze up convolution output size

   ConvTranspose2d with kernel=2, stride=2 doubles spatial size to 128x128, output channels=1.

3. Step 3: Check actual output shape

   Output shape is (batch=1, channels=1, height=128, width=128).

4. Final Answer:

   torch.Size([1, 1, 128, 128]) -> Option B

5. Quick Check:

   Upsample doubles size = 128x128 [OK]

Hint: ConvTranspose2d with stride 2 doubles size [OK]

Common Mistakes:

• Assuming Conv2d changes size without padding
• Ignoring stride effect in ConvTranspose2d
• Mixing channel numbers in output

4. In a U-Net implementation, a student forgot to add skip connections between encoder and decoder layers. What is the most likely effect?

medium

A. The model will not train at all and cause an error

B. The model will have better accuracy due to fewer connections

C. The model will run faster and produce the same results

D. The model will lose fine details and perform worse on segmentation

Solution

1. Step 1: Understand skip connections role

   Skip connections pass fine details from encoder to decoder to improve segmentation accuracy.

2. Step 2: Effect of missing skip connections

   Without them, decoder lacks detail info, causing worse segmentation results.

3. Final Answer:

   The model will lose fine details and perform worse on segmentation -> Option D

4. Quick Check:

   Skip connections = better detail [OK]

Hint: No skip connections = loss of detail [OK]

Common Mistakes:

• Thinking fewer connections always improve accuracy
• Assuming model will error without skip connections
• Believing speed improves without accuracy loss

5. You want to improve a U-Net model's segmentation on small objects in medical images. Which change is most effective?

hard

A. Increase the depth of the encoder to capture more features

B. Remove skip connections to simplify the model

C. Use larger convolution kernels to cover bigger areas

D. Add more pooling layers to reduce spatial size aggressively

Solution

1. Step 1: Identify challenge with small objects

   Small objects need detailed features and context captured by deeper layers.

2. Step 2: Evaluate options

   Increasing encoder depth helps learn more features. Removing skip connections loses detail. Larger kernels may blur small details. More pooling reduces spatial resolution, losing small object info.

3. Final Answer:

   Increase the depth of the encoder to capture more features -> Option A

4. Quick Check:

   Deeper encoder = better small object features [OK]

Hint: Deeper encoder = better small object detection [OK]

Common Mistakes:

• Removing skip connections harms detail
• Larger kernels blur small features
• More pooling loses small object info