The shared audit separates missing candidates from ranking errors. Later ensemble decisions and validation limitations are recorded alongside the missing-annotation diagnostics.
MI-DETR wins 9 classes, RT-DETRv4-X 5, Mr. DETR 1, and Stable-DINO Swin-L 1.
SIZE
RT-DETRv4 leads tiny recall
RT-DETRv4-X leads recall below 0.1% image area; D-FINE-X leads the other major size bands.
DENSITY
MI-DETR leads sparse scenes
MI-DETR leads at 0–5 objects/image; Co-DINO leads the 6–10, 11–20, and >20 slices.
Shared hard-case concentration
Objects <1% · universal failures60.3%
515 of 854
Objects <1% · universal misses78.9%
315 of 399
Objects <1% · validation GT share29.6%
strong overrepresentation in residual errors
854 / 6,449universal failures
399 / 6,449universal misses
140 / 224<0.1% universal failures
124 / 224<0.1% universal misses
A · MISSING-CANDIDATE PROBLEM
Higher resolution → candidate recovery
The audit identified tiny-object misses. MI-DETR1280 improved standalone local and hidden scores, but later ensemble replacement tests did not pass the integration gate.
B · RANKING PROBLEM
Crop rescorer → ranking / precision
External teammate evidence suggests that learned visual crop features can improve the confidence ordering of candidates already proposed by detectors and fusion.
External crop-rescoring evidence · do not compare numerically with our leaderboard
External branch
val_sel
AIdea preliminary
Observed change
Clean EFS WBF
0.7497
0.7175357
reference
EFS + safe ConvNeXt-Tiny crop rescorer
0.7640
approximately 0.729
approximately +0.0143 local / +0.0115 hidden
FEATURE EVIDENCE
Visual crops were necessary
Metadata-only reranking did not provide the same gain; actual crop/image features were required.
STARTING ARCHITECTURE
ConvNeXt-Tiny
ConvNeXt-Base did not beat Tiny in the teammate branch's frozen local gate, so Tiny is the preferred V1 starting point.
COLLABORATION BOUNDARY
Transparent handoff
The repositories document shared artifacts and completed experiments. Exact lineage for the owner-confirmed 0.7910702 submission remains unresolved.
A · MISSING CANDIDATES
Resolution recovers proposals
Completed 1280 controls demonstrate standalone scale gains. Integration tests separately measure whether those recovered candidates improve the protected ensemble.
B · RANKING / PRECISION
Fusion → crop rescoring → LambdaRank
LambdaRank is already documented: 0.7884920 hidden versus 0.7731070 with the same detector/rescorer pipeline. The later scene-context variant was rejected.
High-Confidence FP / Missing-Annotation Audit
Slice
Visual result
Independent support
Interpretation
max any-GT IoU <0.10
120 / 128 visually debris-like
included in 135 / 142 2/3-or-3/3 same-class support
low-GT-overlap evidence
max any-GT IoU <0.30
142 / 150 visually debris-like
120 with 3/3 + 15 with 2/3 same-class support
95.1% independently supported
Detector votes
Co-DINO-1280 · original MI-DETR-1024 · grouped80 Y