RESEARCH UNIT 01.A

VISION & PERCEPTION

Advanced computer vision and neural rendering for autonomous situational awareness.

COMPUTER VISION (CV)

Real-time object detection, classification, and tracking using transformer-based architectures and edge-optimized inference engines.

  • Target Acquisition & Recognition
  • Multi-Object Tracking (MOT)
  • Semantic & Instance Segmentation

NeRFs & GAUSSIAN SPLATTING

Leveraging Neural Radiance Fields and 3D Gaussian Splatting for photorealistic 3D reconstruction from limited 2D imagery.

  • View Synthesis for Mission Planning
  • Digital Twin Generation
  • Real-time Immersive Navigation

SYNTHETIC DATA GEN

Creating high-fidelity, statiscally-aligned synthetic datasets to solve the "data scarcity" problem in military and sensitive domains.

  • GAN and VAE based augmentation
  • Edge-case training for Autonomy
  • Privacy-preserving Data Synthesis

TECHNICAL DEEP DIVE

In 2024, our research focus shifted toward the integration of Generative CV models for real-time target identification. By pre-training on high-fidelity synthetic images, our models achieve 94% accuracy in adversarial edge-case scenarios where traditional training data is unavailable.

Applications include autonomous landing for VTOL systems and high-speed obstacle avoidance in cluttered urban environments.