XAI-based Satellite Image Analysis Acceleration

1. XAI-based collaborative satellite/ground computing acceleration systems and high-precision tracking and image analysis in spatial-temporal domains

Figure 1. Self-Adaptive Learning Model with Imitation Learning

  • To efficiently manage the enormous size of data with increasing satellite image resolution, a collaborative system technology for satellite-to-ground control is developed that combines satellite-on-board runtime acceleration systems with resource management and HPC systems for XAI acceleration processing, which can handle agent models that detect objects of interest.
  • As an XAI technique for the decision-making process of satellite image-based artificial intelligence model, we study the model that explains the results and rationale of image judgment based on artificial intelligence technique to the system operator and develop interactive deep learning system through qualitative XAI analysis.