Pre-trained detection models in Computer Vision - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a pre-trained object detection model from torchvision.

Computer Vision

import torchvision.models as models
model = models.detection.[1](pretrained=True)

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

Afasterrcnn_resnet50_fpn

Bresnet50

Cvgg16

Dmobilenet_v2

Attempts:

3 left

💡 Hint

Common Mistakes

Using classification model names instead of detection model names.

Forgetting to set pretrained=True.

The correct function to load a pre-trained Faster R-CNN detection model is 'fasterrcnn_resnet50_fpn'. Other options are classification models.

2fill in blank

medium

Complete the code to put the model in evaluation mode before inference.

Computer Vision

model = models.detection.fasterrcnn_resnet50_fpn(pretrained=True)
model.[1]()

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

Atrain

Bfit

Ceval

Dpredict

Attempts:

3 left

💡 Hint

Common Mistakes

Using model.train() which sets training mode.

Trying to call model.predict() which is not a PyTorch method.

Calling model.eval() sets the model to evaluation mode, disabling dropout and batch norm updates.

3fill in blank

hard

Fix the error in the code to prepare an input image tensor for the detection model.

Computer Vision

from torchvision import transforms
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[[1], 0.224, 0.225])
])

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

A0.5

B0.229

C1.0

D0.0

Attempts:

3 left

💡 Hint

Common Mistakes

Using incorrect normalization values causing poor model performance.

Mixing mean and std values.

The standard normalization mean and std for ImageNet models use std=0.229 for the red channel.

4fill in blank

hard

Fill both blanks to run inference and get predictions from the model.

Computer Vision

model.eval()
with torch.no_grad():
    predictions = model([[1]])
print(predictions[0][[2]])

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

Ainput_tensor

Bboxes

Clabels

Dscores

Attempts:

3 left

💡 Hint

Common Mistakes

Passing a single tensor instead of a list.

Trying to print 'labels' instead of 'scores' when checking confidence.

The model expects a list of input tensors. The predictions dictionary contains 'boxes', 'labels', and 'scores'. Printing 'scores' shows confidence values.

5fill in blank

hard

Fill all three blanks to filter predictions by confidence threshold and extract bounding boxes.

Computer Vision

threshold = 0.8
pred = predictions[0]
scores = pred[[1]]
boxes = pred[[2]]
filtered_boxes = boxes[scores [3] threshold]

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

Ascores

Bboxes

C>

D<

Attempts:

3 left

💡 Hint

Common Mistakes

Mixing up keys 'scores' and 'boxes'.

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

We get scores and boxes from the prediction dictionary. Filtering boxes where scores are greater than threshold keeps confident detections.