Faster R-CNN usage in PyTorch - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is Faster R-CNN used for in machine learning?

Faster R-CNN is used for object detection, which means it finds and labels objects in images.

beginner

Name the two main parts of Faster R-CNN.

The two main parts are: 1) Region Proposal Network (RPN) that suggests possible object areas, and 2) Fast R-CNN detector that classifies and refines these areas.

intermediate

How do you prepare an image tensor for Faster R-CNN in PyTorch?

You convert the image to a tensor, normalize pixel values, and put it in a list because Faster R-CNN expects a list of images.

intermediate

What does the output of Faster R-CNN look like after prediction?

The output is a list of dictionaries, each with keys like 'boxes' (coordinates), 'labels' (object classes), and 'scores' (confidence).

beginner

Why do we use pretrained models with Faster R-CNN?

Pretrained models have already learned useful features from large datasets, so they help Faster R-CNN detect objects better and train faster.

What does the Region Proposal Network (RPN) in Faster R-CNN do?

A Normalizes image pixels

B Suggests possible object locations in the image

C Classifies objects into categories

D Generates image captions

The RPN suggests regions where objects might be, which are then classified by the next part.

In PyTorch, what input format does Faster R-CNN expect?

A A list of image tensors

B A string path to the image

C A numpy array

D A single image tensor

Faster R-CNN expects a list of image tensors, even if there is only one image.

Which of these is NOT part of Faster R-CNN's output?

A Image captions

B Bounding boxes

C Confidence scores

D Class labels

Faster R-CNN outputs bounding boxes, labels, and scores, but not captions.

Why use a pretrained Faster R-CNN model?

A To generate random images

B To avoid training completely

C To speed up training and improve accuracy

D To reduce image size

Pretrained models help by starting with learned features, making training faster and better.

What library provides a ready Faster R-CNN model in PyTorch?

A matplotlib

B numpy

C pandas

D torchvision

torchvision has pretrained Faster R-CNN models ready to use.

Explain the main steps to use Faster R-CNN for object detection in PyTorch.

Think about loading, preparing input, running prediction, and reading results.

Describe the role of the Region Proposal Network (RPN) and how it helps Faster R-CNN detect objects.

Focus on how RPN narrows down where to look for objects.