Release review
Get setup help

Release review

Two model versions, case by case. Error regressions and unsettled answers flagged; the checks stay with you.

Warehouse perceptionExample data
Local workspace
Currentflow-v2.3
Candidateflow-v2.4
Warehouse regression set12 paired cases · mean endpoint error (px) · 12 with trajectories

Not ready: 3 error regressions, 2 unstable cases that pass on error.

Average case error
3.132.69px-13.9%
CaseError change
5 of 12 casesNeeds a decision first
Regression Unstableaisle-042

Reflective floor

warehousereflection
Current model3.40px
Candidate5.20px
+1.80px
Error by frame px · lower is betterCandidate peak at frame 8
0.05.511.0Frame 1: current 2.92 px, candidate 2.96 pxFrame 2: current 3.06 px, candidate 3.22 pxFrame 3: current 3.23 px, candidate 3.42 pxFrame 4: current 3.60 px, candidate 3.57 pxFrame 5: current 3.67 px, candidate 4.19 pxFrame 6: current 3.33 px, candidate 5.41 pxFrame 7: current 3.88 px, candidate 7.45 pxFrame 8: current 3.94 px, candidate 8.93 pxFrame 9: current 3.50 px, candidate 8.57 pxFrame 10: current 3.20 px, candidate 8.01 pxFrame 11: current 3.60 px, candidate 6.74 pxFrame 12: current 3.43 px, candidate 5.31 pxFrame 13: current 3.20 px, candidate 4.54 pxFrame 14: current 3.54 px, candidate 4.08 pxFrame 15: current 3.26 px, candidate 3.52 pxFrame 16: current 3.03 px, candidate 3.27 px1Frame16
Current modelCandidate

Error increased by 1.80 px, exceeding your 0.3 px limit.

Refinement trajectory update per iteration · px12 iterations
last third0.01.83.6Iteration 1: candidate update 3.258 px, current 2.680 pxIteration 2: candidate update 2.012 px, current 1.697 pxIteration 3: candidate update 1.275 px, current 1.072 pxIteration 4: candidate update 0.850 px, current 0.704 pxIteration 5: candidate update 0.609 px, current 0.469 pxIteration 6: candidate update 0.510 px, current 0.324 pxIteration 7: candidate update 0.439 px, current 0.234 pxIteration 8: candidate update 0.779 px, current 0.168 pxIteration 9: candidate update 1.190 px, current 0.123 pxIteration 10: candidate update 1.303 px, current 0.089 pxIteration 11: candidate update 1.105 px, current 0.067 pxIteration 12: candidate update 0.935 px, current 0.056 px1Iteration12
Current · 4% lateCandidate · 32% late, 3 reversalsreversal: update grew > 5%

Late revision: share of all refinement in the last third of iterations (limit 25%). Reversal: an update larger than the one before it by more than 5% (limit 2).

The candidate made 32% of its refinement in the last third of its iterations and reversed direction 3 times, above your 25% / 2 limits. The current model settled at 4%.The candidate's error rises mid-sequence and its refinement never settles: most of its revisions arrive late. Investigate before release.

Keep this case in your next release check.
Illustrative data · explore the workflowLower is better · equal-weight case averages · stability from label-free trajectories

Make the next comparison yours.

Export once from your evaluator. Import, investigate, and keep checking.