Anchor generator

Your team is developing object detection models based on the Faster R-CNN architecture and using pre-trained backbones. Your task is to create anchor boxes to serve as reference bounding boxes for proposing potential object regions.

You will create 9 standard anchors (3 box sizes and 3 aspect ratios).

torch has been import for you.

This is a part of the course

“Deep Learning for Images with PyTorch”

View Course

Exercise instructions

Import AnchorGenerator from torchvision.models.detection.rpn.
Configure anchor sizes with 3 values: ((32, 64, 128),).
Configure aspect ratio with 3 values `((0.5, 1.0, 2.0),).
Instantiate AnchorGenerator with anchor_sizes and aspect_ratios.

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Import AnchorGenerator
____

# Configure anchor size
anchor_sizes = ____

# Configure aspect ratio
aspect_ratios = ____

# Instantiate AnchorGenerator
rpn_anchor_generator = ____

Edit and Run Code