ORB features in Computer Vision - Model Metrics & Evaluation

Metrics & Evaluation - ORB features

Which metric matters for ORB features and WHY

ORB features are used to find and match key points in images. The main metrics to check are matching accuracy and repeatability. Matching accuracy tells us how many correct matches the ORB detector finds between two images. Repeatability shows if ORB finds the same points when the image changes a bit (like rotation or lighting). These metrics matter because ORB is used in tasks like object recognition or tracking, where correct and stable matches are important.

Confusion matrix or equivalent visualization

For ORB feature matching, we can think of matches as either correct or incorrect. Here is a simple confusion matrix for matches:

      | Predicted Match | Predicted No Match |
      |-----------------|--------------------|
      | True Match (TP) | False Positive (FP) |
      | False Negative (FN)| True Negative (TN)  |
    

For example, if ORB finds 80 correct matches (TP), misses 20 true matches (FN), and finds 10 wrong matches (FP), we can calculate precision and recall to understand quality.

Precision vs Recall tradeoff with examples

Precision means how many of the matches ORB found are actually correct. High precision means fewer wrong matches.

Recall means how many of the true matches ORB was able to find. High recall means ORB finds most of the real matches.

Example: In a robot navigation task, high recall is important so the robot sees enough landmarks to localize itself. But too many wrong matches (low precision) can confuse it. So a balance is needed.

Adjusting ORB parameters (like number of features or matching threshold) changes this tradeoff.

What "good" vs "bad" metric values look like for ORB features

Good values:

• Precision above 0.8 means most matches are correct.
• Recall above 0.7 means most true matches are found.
• Repeatability above 0.75 means ORB finds stable points under changes.

Bad values:

• Precision below 0.5 means many wrong matches, which can cause errors.
• Recall below 0.4 means ORB misses many true points, reducing usefulness.
• Low repeatability means ORB points change a lot with small image changes.

Common pitfalls in ORB feature metrics

• Ignoring false matches: High number of wrong matches can look like good performance if only total matches are counted.
• Overfitting to one image pair: ORB parameters tuned for one pair may not work well on others.
• Data leakage: Using the same images for tuning and testing can give overly optimistic results.
• Ignoring image conditions: ORB performance drops with blur or lighting changes; metrics should test under varied conditions.

Self-check question

Your ORB feature matcher has 90% precision but only 30% recall on a set of image pairs. Is this good for a robot that needs to recognize places reliably? Why or why not?

Answer: This is not good because the low recall (30%) means ORB misses many true matches. The robot may fail to recognize places since it does not find enough correct points, even though the matches it finds are mostly correct (high precision). For reliable recognition, recall should be higher.

Key Result

For ORB features, balancing high precision and recall ensures correct and stable keypoint matches for reliable image tasks.

Practice

1. What is the main purpose of ORB features in computer vision?

easy

A. To find important points and describe them in images

B. To increase the resolution of images

C. To convert images to grayscale

D. To compress images for storage

Solution

1. Step 1: Understand ORB's role

   ORB is designed to detect key points (important points) in images and create descriptors that describe these points.

2. Step 2: Compare options

   The other options describe unrelated image processing tasks, not feature detection and description.

3. Final Answer:

   To find important points and describe them in images -> Option A

4. Quick Check:

   ORB = key points + descriptors [OK]

Hint: Remember ORB finds and describes key points fast [OK]

Common Mistakes:

• Confusing ORB with image enhancement
• Thinking ORB compresses images
• Assuming ORB changes image colors

2. Which of the following is the correct way to create an ORB detector in OpenCV with 500 features?

easy

A. orb = cv2.ORB_create(nfeatures=500)

B. orb = cv2.ORB(500)

C. orb = cv2.createORB(500)

D. orb = cv2.ORB_create(features=500)

Solution

1. Step 1: Recall ORB creation syntax

   The correct OpenCV function to create an ORB detector is cv2.ORB_create(), and the parameter to set number of features is nfeatures.

2. Step 2: Check options

   orb = cv2.ORB_create(nfeatures=500) uses correct function and parameter name. The other options use incorrect function names or parameter names.

3. Final Answer:

   orb = cv2.ORB_create(nfeatures=500) -> Option A

4. Quick Check:

   Use ORB_create with nfeatures [OK]

Hint: Use cv2.ORB_create(nfeatures=...) to set features [OK]

Common Mistakes:

• Using wrong function name like ORB()
• Using incorrect parameter name like features
• Missing parentheses in function call

3. Given the code below, what is the type of the variable kp after running kp, des = orb.detectAndCompute(img, None)?

import cv2
img = cv2.imread('image.jpg', 0)
orb = cv2.ORB_create(nfeatures=1000)
kp, des = orb.detectAndCompute(img, None)

medium

A. A numpy array of descriptors

B. A list of keypoint objects

C. A single keypoint object

D. An integer count of keypoints

Solution

1. Step 1: Understand detectAndCompute output

   The detectAndCompute method returns two values: keypoints and descriptors. Keypoints are returned as a list of keypoint objects.

2. Step 2: Match variable types

   Here, kp receives the keypoints list, des receives the descriptors numpy array.

3. Final Answer:

   A list of keypoint objects -> Option B

4. Quick Check:

   kp = list of keypoints [OK]

Hint: detectAndCompute returns (list, array) [OK]

Common Mistakes:

• Thinking kp is a numpy array
• Assuming kp is a single keypoint
• Confusing descriptors with keypoints

4. What is wrong with this code snippet for detecting ORB features?

import cv2
img = cv2.imread('image.jpg')
orb = cv2.ORB_create(nfeatures=300)
kp, des = orb.detectAndCompute(img, None)
print(len(kp))

medium

A. detectAndCompute requires a mask argument

B. nfeatures parameter is invalid

C. Image is not read in grayscale, causing detectAndCompute to fail

D. print(len(kp)) is incorrect syntax

Solution

1. Step 1: Check image reading mode

   ORB works best with grayscale images. The code reads the image without specifying grayscale, so img is color (3 channels).

2. Step 2: Understand impact on detectAndCompute

   detectAndCompute expects a single channel image; passing a color image can cause incorrect or no keypoints detected.

3. Final Answer:

   Image is not read in grayscale, causing detectAndCompute to fail -> Option C

4. Quick Check:

   Read image with cv2.imread('image.jpg', 0) [OK]

Hint: Always read images in grayscale for ORB [OK]

Common Mistakes:

• Ignoring image color channels
• Thinking nfeatures is invalid
• Assuming mask is mandatory

5. You want to match ORB features between two images but notice very few matches. Which change is most likely to improve the number of good matches?

hard

A. Use a different color space like HSV for detection

B. Decrease the image resolution before detecting features

C. Set the mask parameter to None explicitly

D. Increase the nfeatures parameter when creating the ORB detector

Solution

1. Step 1: Understand nfeatures impact

   nfeatures controls how many keypoints ORB tries to find. Increasing it allows more keypoints to be detected, increasing chances of matches.

2. Step 2: Evaluate other options

   Decreasing resolution reduces detail, hurting matches. Changing color space doesn't affect ORB which works on grayscale. Mask None is default and doesn't affect matches.

3. Final Answer:

   Increase the nfeatures parameter when creating the ORB detector -> Option D

4. Quick Check:

   More features = more matches [OK]

Hint: More features means more chances to match [OK]

Common Mistakes:

• Reducing image size to get more features
• Changing color space for ORB detection
• Misunderstanding mask parameter role