ANIMA SLOUGHI — Wave 10 WARDOG

Paper: Similarity-Guided Layer-Adaptive ViT for UAV Tracking (SGLATrack) ArXiv: https://arxiv.org/abs/2503.06625 GitHub: https://github.com/GXNU-ZhongLab/SGLATrack Defense Score: 40/50 | Tier: T3 Wave: 10 — WARDOG (War Dog Breeds) Focus: UAV/Drone Defense for Shenzhen Robot Fair

Overview

Similarity-guided adaptive layer pruning for efficient ViT-based UAV tracking

Key Result: SOTA real-time with dynamic layer pruning

Quick Start

# Install dependencies
uv pip install -e ".[dev]"

# Run synthetic inference
python -m anima_sloughi infer --backend auto --synthetic --with-yolo26-prior

# Run synthetic training/evaluation
python -m anima_sloughi train --backend cpu --steps 3 --batch-size 2
python -m anima_sloughi evaluate --backend cpu --dataset synthetic

# If package is not installed in editable mode:
PYTHONPATH=src python -m anima_sloughi infer --backend cpu --synthetic

Project Structure

project_sloughi/
├── src/anima_sloughi/   # Source code
├── tests/                   # Unit tests
├── configs/                 # Configuration files
├── scripts/                 # Utility scripts
├── papers/                  # Paper PDF
├── docker/                  # Docker setup
├── CLAUDE.md               # Agent instructions
├── PRD.md                  # Production requirements
├── NEXT_STEPS.md           # Execution ledger
└── MODULE_TODO.md          # Implementation checklist

Dual Compute

All code runs on both MLX (Apple Silicon) and CUDA (GPU server). See src/anima_sloughi/device.py for the abstraction layer.

License

Research use only. See paper for original license terms.