Intersection over Union (IoU) for Tracking

mediumPROObject TrackingTrackingCV: Motion Estimation

Implement IoU computation for bounding box overlap in object tracking. This task involves calculating the overlap between two bounding boxes, a crucial step in evaluating the performance of object tracking algorithms.

The concept of Intersection over Union (IoU) is fundamental in computer vision, particularly in object tracking and detection. It measures the overlap between two bounding boxes, $A$ and $B$ , and is defined as the ratio of the intersection area to the union area. The intersection area represents the common region between the two boxes, while the union area represents the total area covered by both boxes.

To compute IoU, we need to follow these steps:

Calculate the intersection area by finding the overlapping region between the two boxes.
Calculate the individual areas of the two boxes.
Calculate the union area by adding the individual areas and subtracting the intersection area.

IoU = \frac{|A \cap B|}{|A \cup B|} = \frac{\text{Intersection Area}}{\text{Union Area}}

This technique is widely used in object tracking evaluation metrics, such as the MOTA metric.

Example

Input

box1 = [0, 0, 10, 10]
box2 = [5, 5, 15, 15]

Output

0.143

Reasoning

Inputs: each box is $[x_1, y_1, x_2, y_2]$ , its top-left corner $(x_1, y_1)$ and bottom-right corner $(x_2, y_2)$ . Here box1 covers $x, y \in [0, 10]$ and box2 covers $x, y \in [5, 15]$ .
Box areas: width times height, so box1 has $(10-0)(10-0) = 100$ and box2 has $(15-5)(15-5) = 100$ .
Intersection: the region inside both boxes runs from $x = 5$ to $x = 10$ and from $y = 5$ to $y = 10$ , a $5 \times 5$ square with area $25$ .
Union: adding the two areas counts that shared square twice, so $|A \cup B| = 100 + 100 - 25 = 175$ .
Output: $\text{IoU} = \frac{25}{175} = \frac{1}{7} \approx 0.142857$ , shown as 0.143. The function returns that unrounded ratio as a single float in $[0, 1]$ .
Notice that shifting a box diagonally by half its side length in both $x$ and $y$ leaves an IoU of only $\frac{1}{7}$ , not $\frac{1}{2}$ : IoU punishes misalignment hard. Identical boxes score $1$ ; boxes that share no area score $0$ .

Constraints

box1, box2: Bounding boxes as [x1, y1, x2, y2]
Return: IoU score (0-1)

Learn this in

Motion EstimationCV Study Plan

More from CV: Motion Estimation

Hungarian Algorithm for MOThard Rolling Shutter Correctionhard Horn-Schunck Global Optical Flowmedium Bilinear Flow Warpinghard Trajectory Smoothing Filtermedium

Concept primerPREVIEW

Intersection over Union (IoU) for Tracking: Background & Strategy

Background Knowledge

What is IoU and Why It Matters

Intersection over Union is a fundamental metric in computer vision that quantifies the spatial overlap between two bounding boxes. The formula divides the area where two boxes overlap by the total area they collectively cover. In object tracking, IoU serves multiple critical purposes: it helps determine whether a detection in the current frame corresponds to an object tracked in the previous frame, it's used in non-maximum suppression (NMS) to eliminate redundant detections, and it forms the basis for data association in multi-object tracking systems. Values range from 0 (no overlap) to 1 (perfect overlap).

Bounding Box Representation

Bounding boxes are typically represented in one of two formats: (1) corner coordinates as (x₁, y₁, x₂, y₂) where (x₁, y₁) is the top-left corner and (x₂, y₂) is the bottom-right corner, or (2) center-based as (cx, cy, w, h) where cx and cy are the center coordinates and w, h are width and height. Understanding format conversion is essential because you may receive boxes in either format. The choice affects how you compute intersection and union areas—corner format makes intersection computation more straightforward since you can directly compare coordinates.

Why IoU Matters for Tracking

In multi-object tracking, IoU enables the data association problem: matching detections across consecutive frames to maintain consistent object identities. When objects are close or overlapping, naive distance-based matching fails; IoU provides a more robust spatial similarity measure. The metric is also used in evaluation metrics like MOTA (Multiple Object Tracking Accuracy) and helps reduce identity switches—a critical failure mode where the same object is assigned different IDs in consecutive frames.

Algorithm/Approach

The general approach to computing IoU between two bounding boxes follows this pattern:

Normalize input: Ensure both boxes are in the same coordinate format (typically corner-based for easier computation)
Compute intersection: Find the overlapping rectangular region between the two boxes
Compute union: Calculate the total area covered by both boxes
Calculate ratio: Divide intersection by union

The key insight is that intersection and union can be computed using simple coordinate comparisons rather than pixel-level operations, making the algorithm efficient even for large-scale tracking scenarios.

Step-by-Step Strategy

Step 1: Parse and Validate Input

Accept two bounding boxes in your chosen format
Validate that coordinates are valid (e.g., x₁ < x₂ for corner format)
Consider edge cases: zero-area boxes, identical boxes, non-overlapping boxes

PRO

Continue the full explanation

You're reading the free preview. Unlock the complete walkthrough, the code editor, test runner and reference solution with Premium.

PRO

Editor locked

The code editor is locked for Pro problems. It is only available for free problems. Please upgrade to gain access to the code editor for all problems.

solution.py

Python 3.13

Test results