EdgefieldConsulting

Services
Quiz
Offers
Book
Work
About
Contact

← All checklists

Edge AI & Computer Vision · Free resource

Real-time YOLO Vision Pipeline Checklist

Two-stage detection, tracking, and georeferencing that hold up on Jetson Orin — from dataset hygiene to GPS waypoint injection.

0/16 · 0%

Request an engagement

1. Data & evaluation

Train / val / test splits fixed; no leakage from flight videos
Class taxonomy matches ops language (not researcher nicknames only)
Confusion matrix reviewed for costly false positives in the field
Hard negatives from glare, motion blur, and distant objects included

2. Detection pipeline

Detector (YOLO family or equivalent) latency fits frame budget on Orin
NMS / confidence thresholds tuned on field footage, not only COCO
Second-stage refine (crop / classifier) justified if used — cost measured
Tracking IDs stable across short occlusions; ID-switch cost understood

3. Georeferencing & waypoints

Pixel → ground projection uses current attitude, altitude, and camera model
GPS / EKF pose quality gated before injecting waypoints
Waypoint rate limited; no setpoint spam on every detection frame
Human confirmation or geofence required for high-risk target classes

4. Ops integration

Overlay / HUD shows confidence and pose quality to the operator
Detection events logged with pose for post-mission audit
Privacy / ROE constraints documented for recording and retention
Fallback: operator can disable auto-waypoint with one clear switch

Want this applied to your stack? Book a discovery call or request an async audit at edgefield-consulting.pages.dev · aykumar2106@gmail.com

© Edgefield Consulting — Software engineered for the field. Research that ships.

Prefer a call? Open Calendly

Edgefield

Consulting

Software engineered for the field. Research that ships.

Services Quiz Offers Book Work About Contact GitHub

© 2026 Edgefield Consulting

Taiwan · India · Remote