Human pose estimation concept in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a pre-trained human pose estimation model using OpenPose.

Computer Vision

import cv2
net = cv2.dnn.readNetFromCaffe('pose_deploy_linevec.prototxt', '[1]')

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

Apose_iter_100000.caffemodel

Bpose_iter_440000.caffemodel

Cpose_deploy_linevec.prototxt

Dpose_deploy.prototxt

Attempts:

3 left

💡 Hint

Common Mistakes

Using the prototxt file instead of the caffemodel file for weights.

Confusing the model structure file with the weights file.

The model weights file 'pose_iter_440000.caffemodel' is loaded with the prototxt to create the network.

2fill in blank

medium

Complete the code to prepare the input image blob for the pose estimation model.

Computer Vision

blob = cv2.dnn.blobFromImage(image, 1.0 / 255, (368, 368), (0, 0, 0), swapRB=True, crop=False)
net.setInput([1])

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

Ablob

Bimage

Cnet

Dinput

Attempts:

3 left

💡 Hint

Common Mistakes

Passing the original image instead of the blob to the network.

Passing the network object itself as input.

The blob created from the image is set as input to the network.

3fill in blank

hard

Fix the error in the code to extract the keypoints from the model output heatmaps.

Computer Vision

output = net.forward()
H, W = output.shape[2], output.shape[3]
points = []
for i in range(num_keypoints):
    probMap = output[0, i, :, :]
    minVal, prob, minLoc, point = cv2.minMaxLoc(probMap)
    x = (image.shape[1] * point[0]) / [1]
    y = (image.shape[0] * point[1]) / H
    if prob > threshold:
        points.append((int(x), int(y)))
    else:
        points.append(None)

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

AW

Cnum_keypoints

Dthreshold

Attempts:

3 left

💡 Hint

Common Mistakes

Mixing up height and width when scaling coordinates.

Using the wrong dimension for scaling x or y.

The height H should be used to scale the y-coordinate, and width W for x-coordinate. Here, x is scaled by width, so the denominator must be width W, but the code mistakenly uses H. The blank is for the denominator of x coordinate, which should be W.

4fill in blank

hard

Fill both blanks to create a dictionary of keypoints with their names and coordinates.

Computer Vision

keypoints_dict = [1](name: points[i] for i, name in [2](keypoint_names))

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

Adict

Blist

Cenumerate

Drange

Attempts:

3 left

💡 Hint

Common Mistakes

Using list() instead of dict() for dictionary creation.

Using range() instead of enumerate() to get index and name.

To create a dictionary from keypoint names and points, use dict() and enumerate() to get index and name pairs.

5fill in blank

hard

Fill all three blanks to filter detected keypoints with confidence above threshold and create a list of their coordinates.

Computer Vision

filtered_points = [points[i] for i, prob in [1](output[0, :, :, :].max(axis=(1, 2))) if prob [2] threshold and points[i] is not [3]]

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

Aenumerate

B>

CNone

Drange

Attempts:

3 left

💡 Hint

Common Mistakes

Using range instead of enumerate for index and value pairs.

Using '<' instead of '>' for filtering probabilities.

Checking points[i] against wrong value instead of None.

Use enumerate to get index and probability, filter where prob > threshold and point is not None.