Face detection with deep learning in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load the pre-trained face detection model.

Computer Vision

net = cv2.dnn.readNetFromCaffe('[1]', 'res10_300x300_ssd_iter_140000.caffemodel')

Drag options to blanks, or click blank then click option'

Aconfig.pbtxt

Bdeploy.prototxt.txt

Cface_detector.prototxt

Dmodel.caffemodel

Attempts:

3 left

💡 Hint

Common Mistakes

Using the weights file instead of the architecture file.

Using a TensorFlow config file instead of Caffe prototxt.

The readNetFromCaffe function requires the model architecture file, which is deploy.prototxt.txt.

2fill in blank

medium

Complete the code to prepare the input blob for the face detector.

Computer Vision

blob = cv2.dnn.blobFromImage(image, 1.0, ([1], [1]), (104.0, 177.0, 123.0))

Drag options to blanks, or click blank then click option'

A300

B224

C128

D256

Attempts:

3 left

💡 Hint

Common Mistakes

Using 224 which is common for classification but not this detector.

Using different width and height values.

The face detector expects input images resized to 300x300 pixels.

3fill in blank

hard

Fix the error in the code to extract confidence scores from detections.

Computer Vision

confidence = detections[0, 0, i, [1]]

Drag options to blanks, or click blank then click option'

A2

B4

C3

D1

Attempts:

3 left

💡 Hint

Common Mistakes

Using index 1 which is class id.

Using index 3 or 4 which are bounding box coordinates.

The confidence score is at index 2 in the detection array.

4fill in blank

hard

Fill both blanks to compute bounding box coordinates from detection.

Computer Vision

box = detections[0, 0, i, 3:7] * np.array([[1], [2], [1], [2]])

Drag options to blanks, or click blank then click option'

Aw

Bh

Cimage.shape[1]

Dimage.shape[0]

Attempts:

3 left

💡 Hint

Common Mistakes

Using width for both x and y coordinates.

Using height for both x and y coordinates.

The bounding box coordinates are normalized, so multiply by image width and height respectively.

5fill in blank

hard

Fill all three blanks to draw rectangles around detected faces with confidence above threshold.

Computer Vision

if confidence > [1]:
    (startX, startY, endX, endY) = box.astype(int)
    cv2.rectangle(image, (startX, startY), (endX, endY), [2], [3])

Drag options to blanks, or click blank then click option'

A0.5

B(0, 255, 0)

C2

D0.75

Attempts:

3 left

💡 Hint

Common Mistakes

Using a confidence threshold too low or too high.

Using wrong color format or thickness type.

We draw rectangles for detections with confidence above 0.5, using green color and thickness 2.