Autonomy Failed. We Fixed It. Victus builds contested autonomy software that works where GPS doesn’t, on any platform, from the seafloor to low Earth orbit.
getvictus.ai
info@getvictus.ai
© 2026 victus technologies
All rights reserved
Datasheet
PhantomSIGHT™
OPERATOR-TRAINED TARGET DETECTION
Field-trainable Automatic Target Recognition (ATR) for unmanned systems. From operator-supplied imagery to a deployed detection model in hours — running on Commercial Off-The-Shelf (COTS) edge compute. No specialized machine-learning team required, no cloud dependency, no model black box. Group 1–3 software deployment for Unmanned Aerial Systems (UAS), ground robotics, and surface and sub-surface vessels.
Core Capabilities
FEW-SHOT FINE-TUNING
Mission-relevant detection from as few as five labeled images.
AUTOMATED LABEL TRACKING
Operator-defined target labels propagate across frames — train to specific subjects of interest, not pixels.
DETECTION STABILIZATION
Multi-stage temporal filtering eliminates frame flicker.
PLUGIN AUGMENTATION
Geometric, environmental, camouflage, and copy-paste strategies.
OPERATOR-LED WORKFLOW
Annotate, train, and deploy from one browser interface.
AIR-GAPPED CAPABLE
Full pipeline runs without external network connectivity.
Operator Pipeline
Step 1
Capture
Any sensor, any format. Video clips, single frames, or both.
Step 2
Annotate
Box the target once. Tracker propagates the label downstream.
Step 3
Train
Launch fine-tune from the browser. Live metrics, live convergence.
Step 4
Validate
Held-out frame scoring. Accuracy and stability before fielding.
Step 5
Deploy
Compile to ONNX or TensorRT. Drop onto edge inference compute.