Face detection with deep learning in Computer Vision - Model Metrics & Evaluation

Metrics & Evaluation - Face detection with deep learning

Which metric matters for face detection and WHY

For face detection, the key metrics are Precision and Recall. Precision tells us how many detected faces are actually faces, so it helps avoid false alarms. Recall tells us how many real faces the model finds, so it helps avoid missing faces. Both matter because missing a face or wrongly detecting a face can cause problems in real life, like security or photo tagging.

Confusion matrix for face detection

      | Predicted Face | Predicted No Face |
      |----------------|-------------------|
      | True Positive  | False Positive    |
      | False Negative | True Negative     |

      Example:
      TP = 90 (faces correctly detected)
      FP = 10 (wrongly detected faces)
      FN = 15 (faces missed)
      TN = 885 (correctly ignored non-faces)

      Total samples = 90 + 10 + 15 + 885 = 1000
    

Precision vs Recall tradeoff with examples

If the model tries to find every face (high recall), it might detect some things that are not faces (lower precision). For example, in a security camera, missing a face (low recall) is bad because you want to catch everyone. But if it detects too many false faces (low precision), it wastes time checking wrong alerts.

On the other hand, if the model is very strict and only detects faces it is sure about (high precision), it might miss some faces (low recall). This might be okay for photo apps where false faces annoy users, but not for safety systems.

What good vs bad metric values look like for face detection

Good: Precision and Recall both above 90%. This means most detected faces are real, and most real faces are found.

Bad: Precision below 70% means many false alarms. Recall below 70% means many faces missed. For example, 50% recall means half the faces are not detected, which is usually unacceptable.

Common pitfalls in face detection metrics

• Accuracy paradox: If most images have no faces, a model that always says "no face" can have high accuracy but is useless.
• Data leakage: Testing on images very similar to training can give too optimistic metrics.
• Overfitting: Very high training metrics but poor test metrics mean the model learned noise, not real face patterns.

Self-check question

Your face detection model has 98% accuracy but only 12% recall on faces. Is it good for production? Why or why not?

Answer: No, it is not good. The high accuracy is misleading because most images might not have faces. The very low recall means the model misses almost all faces, which defeats the purpose of face detection.

Key Result

Precision and recall are key for face detection; both must be high to ensure real faces are found without many false alarms.

Practice

1. What is the main purpose of face detection in deep learning?

easy

A. To find and locate faces in images or videos

B. To recognize the identity of a person

C. To generate new face images

D. To remove faces from images

Solution

1. Step 1: Understand face detection

   Face detection means finding where faces are in an image, not who they belong to.

2. Step 2: Differentiate from face recognition

   Face recognition identifies who the face belongs to, which is different from detection.

3. Final Answer:

   To find and locate faces in images or videos -> Option A

4. Quick Check:

   Face detection = Find faces [OK]

Hint: Face detection locates faces, recognition identifies them [OK]

Common Mistakes:

• Confusing detection with recognition
• Thinking detection creates new images
• Assuming detection removes faces

2. Which of the following is the correct way to load a Haar Cascade face detector in OpenCV (Python)?

easy

A. face_cascade = cv2.FaceDetector('haarcascade_frontalface_default.xml')

B. face_cascade = cv2.loadCascade('haarcascade_frontalface_default.xml')

C. face_cascade = cv2.Cascade('haarcascade_frontalface_default.xml')

D. face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')

Solution

1. Step 1: Recall OpenCV syntax for Haar Cascade

   OpenCV uses cv2.CascadeClassifier to load Haar Cascade XML files.

2. Step 2: Check each option's method name

   Only cv2.CascadeClassifier is correct; others are invalid or do not exist.

3. Final Answer:

   face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml') -> Option D

4. Quick Check:

   Use CascadeClassifier to load Haar cascades [OK]

Hint: Use CascadeClassifier to load Haar cascades in OpenCV [OK]

Common Mistakes:

• Using wrong method names like loadCascade or Cascade
• Confusing face detector with face recognizer
• Missing the .xml file extension

3. Given this code snippet, what will be the output if the image contains two faces?

import cv2
face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
img = cv2.imread('group_photo.jpg')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5)
print(len(faces))

medium

A. 2

B. 1

C. 0

D. Error

Solution

1. Step 1: Understand detectMultiScale output

   detectMultiScale returns a list of rectangles for each detected face.

2. Step 2: Count detected faces in the image

   Since the image has two faces, len(faces) will be 2.

3. Final Answer:

   2 -> Option A

4. Quick Check:

   Faces detected = 2 [OK]

Hint: detectMultiScale returns all faces found; count them [OK]

Common Mistakes:

• Assuming detectMultiScale returns a boolean
• Confusing number of faces with confidence scores
• Forgetting to convert image to grayscale

4. What is wrong with this face detection code snippet?

import cv2
face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
img = cv2.imread('photo.jpg')
faces = face_cascade.detectMultiScale(img)
print(len(faces))

medium

A. detectMultiScale requires a color image

B. The XML file path is incorrect

C. The image should be converted to grayscale before detection

D. print(len(faces)) should be print(faces)

Solution

1. Step 1: Check input type for detectMultiScale

   detectMultiScale expects a grayscale image, not a color image.

2. Step 2: Identify missing grayscale conversion

   The code misses converting img to grayscale before detection.

3. Final Answer:

   The image should be converted to grayscale before detection -> Option C

4. Quick Check:

   detectMultiScale needs grayscale input [OK]

Hint: Always convert images to grayscale before face detection [OK]

Common Mistakes:

• Passing color images directly to detectMultiScale
• Assuming detectMultiScale works on any image format
• Confusing print(len(faces)) with print(faces)

5. You want to improve face detection accuracy in a crowded photo using Haar Cascades. Which approach is best?

hard

A. Increase minNeighbors parameter to reduce false positives

B. Decrease scaleFactor to detect smaller faces more precisely

C. Use color images instead of grayscale for detection

D. Skip preprocessing and use raw images directly

Solution

1. Step 1: Understand scaleFactor effect

   Lower scaleFactor means the detector checks more scales, helping find smaller faces.

2. Step 2: Compare with other options

   Increasing minNeighbors reduces false positives but may miss faces; color images and skipping preprocessing reduce accuracy.

3. Final Answer:

   Decrease scaleFactor to detect smaller faces more precisely -> Option B

4. Quick Check:

   Lower scaleFactor = better small face detection [OK]

Hint: Lower scaleFactor to catch smaller faces in crowded images [OK]

Common Mistakes:

• Increasing minNeighbors too much misses faces
• Using color images instead of grayscale
• Skipping image preprocessing steps