Computer Vision Pipeline

[miyatakazuya]

🚀 Feature

Set up the CV Pipeline to intake images from the PyCam Client to perform object detection and localization of ground targets.

Motivation

New Rule Changes that call for new detection methods, as well as the integration of a new camera.

Pitch

Inputs:

• Full resolution image (after any image-processing)
• Plane positional data from the time the picture was taken
• Camera Gimbal data?

Components

• Object Detection / Classification
  • Input: Full resolution image (after any image-processing)
  • Detection model identifies potential objects, annotating their pixel position.
  • Keep track of 5 unique object types.
  • Output:: List of cropped targets where each one is annotated with their pixel position within the original image
• Localization
  Input: Position of target in image space, plane positional data
  Output: Latitude/Longitude of specified target