Skip to content

Latest commit

 

History

History
26 lines (20 loc) · 985 Bytes

File metadata and controls

26 lines (20 loc) · 985 Bytes

Constellation Identification

🏗️ Work In Progress

This project simulates star constellations and uses a Graph Attention Network (GAT) to predict the rotational angles (u, v) of a robot's head from a graph of the captured image.

Objectives

  • Simulate star dataset:
    • Simulate stars
    • Convert stars into graph representations (stars as nodes, distances as edges)
    • Convert stars into image samples
    • Convert images into graph representations (stars as nodes, distances as edges)
  • Implement GAT model:
    • Predict the true position of stars and/or the camera angle ((u), (v)).
    • Should we predict star positions and deduce the camera angle, or predict the camera angle directly?

Execution

make dummy_data # simulate dummy data (naive)
make clean # clean

# not implemented yet
make simulate_data # simulate stars positions => images of stars => graph of stars => prediction
make train # train the GAT