YOLO architecture concept in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to define the input shape for the YOLO model.

Computer Vision

input_shape = ([1], [2], 3)

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

A416

B224

C512

D128

Attempts:

3 left

💡 Hint

Common Mistakes

Using non-square input shapes like 224x416.

Forgetting the 3 color channels for RGB images.

The YOLO model commonly uses 416x416 pixel input images with 3 color channels (RGB).

2fill in blank

medium

Complete the code to create a convolutional layer in YOLO with 32 filters.

Computer Vision

conv_layer = Conv2D([1], (3, 3), padding='same', activation='leaky_relu')

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

A32

B16

C64

D128

Attempts:

3 left

💡 Hint

Common Mistakes

Using too few or too many filters like 16 or 128 at the start.

Forgetting the activation function.

YOLO starts with a convolutional layer of 32 filters to extract features.

3fill in blank

hard

Fix the error in the YOLO output layer code to predict bounding boxes.

Computer Vision

output = Conv2D([1], (1, 1), activation='linear')(previous_layer)

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

A255

B5

C3

D85

Attempts:

3 left

💡 Hint

Common Mistakes

Using 85 which is values per bounding box, but the layer outputs 255 for 3 boxes.

Using 5 which is for one box (4 coords + 1 objectness) without classes.

The output layer predicts 255 values per grid cell for 3 bounding boxes: 3 × (4 box coords + 1 objectness + 80 class probabilities) = 255.

4fill in blank

hard

Fill both blanks to complete the YOLO loss function components.

Computer Vision

loss = [1] + [2]

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

Alocalization_loss

Bclassification_loss

Cregularization_loss

Dreconstruction_loss

Attempts:

3 left

💡 Hint

Common Mistakes

Confusing reconstruction loss with YOLO loss.

Ignoring classification loss in total loss.

YOLO loss combines localization loss (for bounding box accuracy) and classification loss (for object class prediction).

5fill in blank

hard

Fill all three blanks to complete the YOLO detection pipeline steps.

Computer Vision

detections = [1](image)
filtered = [2](detections, threshold=0.5)
final_boxes = [3](filtered)

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

Amodel.predict

Bnon_max_suppression

Cscale_boxes

Dimage_preprocessing

Attempts:

3 left

💡 Hint

Common Mistakes

Applying image preprocessing after prediction.

Skipping non-max suppression step.

YOLO pipeline: predict boxes, apply non-max suppression to remove overlaps, then scale boxes to original image size.