Strategy / Research priority

Revisit / Graveyard

This 1–10 rank is a research-priority ranking, not a quality score. 1 means revisit seriously; 10 means dead unless fundamentally reformulated.

Worth Revisiting · 1–3 · Deserves or may deserve another serious look
1
DINOv3 appearance embeddings
METHOD-DINO-V3
Why parked / rejected
Active candidate; priority reflects validation urgency.
Evidence
EVIDENCE-2026-09-001
What was tested
Daylight football, fixed camera, 1080p
What was NOT tested
Night, rain, moving cameras
Trigger conditions to revisit
Broader labelled session set
Current status
Promising
1
Relational topology
METHOD-TOPOLOGY-V1
Why parked / rejected
Active candidate; priority reflects validation urgency.
Evidence
EVIDENCE-2026-09-003
What was tested
Open play and short occlusions
What was NOT tested
Corners, substitutions, goalkeeper transitions
Trigger conditions to revisit
Validate on Session B ground truth
Current status
Promising
Parked · 4–6 · Hold until evidence or dependencies change
5
Colour histogram continuity
METHOD-COLOUR-HIST
Why parked / rejected
Insufficient discrimination within same team.
Evidence
EVIDENCE-2026-09-002
What was tested
One daylight session; HSV 32-bin histograms
What was NOT tested
Adaptive white balance; regional histograms on 4K
Trigger conditions to revisit
Revisit as a low-weight fusion feature after colour calibration
Current status
Parked
Near-Dead · 7–8 · Low priority without a strong trigger
Dead Unless Reformulated · 9–10 · Effectively dead unless fundamentally reformulated
9
Gait signature
METHOD-GAIT-V1
Why parked / rejected
Signal resolution was inadequate in the tested wide-angle footage; this does not invalidate close-camera gait.
Evidence
No linked artifact
What was tested
1080p wide camera, 2–5 second tracks, 12 players
What was NOT tested
4K crops, close camera, longer uninterrupted tracks
Trigger conditions to revisit
Fundamentally reformulate using 4K or a dedicated close camera
Current status
Rejected