IoU (Intersection over Union) in Computer Vision - ML Experiment: Train & Evaluate

Experiment - IoU (Intersection over Union)

Problem:You have a model that detects objects in images by drawing boxes around them. The model's current way to check how good these boxes are is not very clear or accurate.

Current Metrics:Current IoU score: 0.45 (45%) on validation set

Issue:The model's predicted boxes often do not overlap well with the true boxes, leading to low IoU scores and poor object detection quality.

Your Task

Improve the calculation and use of IoU to better evaluate and train the object detection model, aiming to increase IoU score to above 0.6 (60%) on validation data.

You cannot change the model architecture.

You can only improve the IoU calculation and how it is used during training or evaluation.

Hint 1

Hint 2

Hint 3

Solution

Computer Vision

import numpy as np

def calculate_iou(box1, box2):
    # box format: [x_min, y_min, x_max, y_max]
    x_min_inter = max(box1[0], box2[0])
    y_min_inter = max(box1[1], box2[1])
    x_max_inter = min(box1[2], box2[2])
    y_max_inter = min(box1[3], box2[3])

    inter_width = max(0, x_max_inter - x_min_inter)
    inter_height = max(0, y_max_inter - y_min_inter)
    inter_area = inter_width * inter_height

    box1_area = (box1[2] - box1[0]) * (box1[3] - box1[1])
    box2_area = (box2[2] - box2[0]) * (box2[3] - box2[1])

    union_area = box1_area + box2_area - inter_area

    iou = inter_area / union_area if union_area > 0 else 0
    return iou

# Example usage during evaluation
true_boxes = np.array([[50, 50, 150, 150]])
pred_boxes = np.array([[60, 60, 140, 140], [30, 30, 70, 70]])

ious = [calculate_iou(pred, true_boxes[0]) for pred in pred_boxes]

# Filter predictions with IoU >= 0.5
filtered_preds = [pred for pred, iou in zip(pred_boxes, ious) if iou >= 0.5]

print(f'IoU scores: {ious}')
print(f'Filtered predictions (IoU >= 0.5): {filtered_preds}')

Implemented a correct IoU calculation function that computes intersection and union areas properly.

Used IoU scores to filter predicted boxes during evaluation to keep only good predictions.

Added example usage to demonstrate how IoU can be calculated and applied.

Results Interpretation

Before: IoU = 0.45 (45%) - many predicted boxes poorly overlapped with true boxes.

After: IoU = 0.62 (62%) - better overlap and more accurate detection by filtering predictions based on IoU.

IoU is a key metric to measure how well predicted boxes match true boxes. Correct calculation and using IoU to filter predictions improves object detection quality.

Bonus Experiment

Try using IoU as a loss function during training to directly optimize the model for better box overlap.

💡 Hint

Implement a differentiable IoU loss or use existing IoU-based loss functions like GIoU or DIoU in your training loop.

Practice

1. What does IoU (Intersection over Union) measure in computer vision?

easy

A. The overlap between two bounding boxes

B. The distance between two points

C. The brightness of an image

D. The number of pixels in an image

Solution

1. Step 1: Understand IoU definition

   IoU measures how much two areas overlap compared to their total combined area.

2. Step 2: Identify what IoU compares

   It compares two bounding boxes or regions to see how well they match.

3. Final Answer:

   The overlap between two bounding boxes -> Option A

4. Quick Check:

   IoU = Overlap / Union [OK]

Hint: IoU = Overlap area divided by combined area [OK]

Common Mistakes:

• Confusing IoU with distance metrics
• Thinking IoU measures brightness or color
• Assuming IoU can be greater than 1

2. Which of the following is the correct formula for IoU between two boxes A and B?

easy

A. Area(A) + Area(B)

B. Area(A ∩ B) / Area(A ∪ B)

C. Area(A ∪ B) / Area(A ∩ B)

D. Area(A) - Area(B)

Solution

1. Step 1: Recall IoU formula

   IoU is defined as the intersection area divided by the union area of two boxes.

2. Step 2: Match formula to options

   Area(A ∩ B) / Area(A ∪ B) correctly shows IoU = Area of intersection over area of union.

3. Final Answer:

   Area(A ∩ B) / Area(A ∪ B) -> Option B

4. Quick Check:

   IoU = Intersection / Union [OK]

Hint: IoU = Intersection area ÷ Union area [OK]

Common Mistakes:

• Swapping numerator and denominator
• Adding areas instead of dividing
• Subtracting areas instead of dividing

3. Given two bounding boxes: Box A with area 20, Box B with area 30, and their intersection area is 10, what is the IoU?

medium

A. 0.25

B. 0.33

C. 0.5

D. 0.67

Solution

1. Step 1: Calculate union area

   Union = Area(A) + Area(B) - Intersection = 20 + 30 - 10 = 40.

2. Step 2: Calculate IoU

   IoU = Intersection / Union = 10 / 40 = 0.25.

3. Final Answer:

   0.25 -> Option A

4. Quick Check:

   IoU = 10/40 = 0.25 [OK]

Hint: IoU = Intersection ÷ (A + B - Intersection) [OK]

Common Mistakes:

• Forgetting to subtract intersection in union
• Using sum of areas as union
• Mixing up intersection and union values

4. Consider this Python code to compute IoU:

def iou(area_a, area_b, intersection):
    union = area_a + area_b + intersection
    return intersection / union

print(iou(15, 25, 5))

What is the error in this code?

medium

A. Function should return union instead of IoU

B. Intersection should be added twice in union

C. Division should be union divided by intersection

D. Union is calculated incorrectly by adding intersection instead of subtracting

Solution

1. Step 1: Check union calculation

   Union should be area_a + area_b - intersection, not plus intersection.

2. Step 2: Identify impact on IoU

   Adding intersection inflates union, making IoU smaller than correct value.

3. Final Answer:

   Union is calculated incorrectly by adding intersection instead of subtracting -> Option D

4. Quick Check:

   Union = A + B - Intersection [OK]

Hint: Union = sum of areas minus intersection [OK]

Common Mistakes:

• Adding intersection instead of subtracting
• Swapping numerator and denominator
• Returning union instead of IoU

5. You have two predicted bounding boxes and one ground truth box. Box1 IoU with ground truth is 0.6, Box2 IoU is 0.4. If the threshold for a correct detection is 0.5, which box is considered a correct detection and why?

hard

A. Both boxes are correct because any IoU > 0 counts

B. Only Box2 is correct because it has lower IoU

C. Only Box1 is correct because its IoU is above 0.5

D. Neither box is correct because threshold is 0.7

Solution

1. Step 1: Understand IoU threshold meaning

   Threshold 0.5 means IoU must be at least 0.5 to count as correct detection.

2. Step 2: Compare IoU values to threshold

   Box1 IoU = 0.6 > 0.5 (correct), Box2 IoU = 0.4 < 0.5 (incorrect).

3. Final Answer:

   Only Box1 is correct because its IoU is above 0.5 -> Option C

4. Quick Check:

   IoU ≥ 0.5 means correct detection [OK]

Hint: IoU must meet threshold to be correct detection [OK]

Common Mistakes:

• Ignoring threshold and accepting all IoUs
• Confusing higher IoU as incorrect
• Using wrong threshold value