feat(agent): dedupe near-duplicate crops before the SigLIP embed
Figure boxes are already NMS-merged (iou 0.6) and each YOLO detector self-NMSes, but the combined per-frame crop pile (figure→concept ∪ anatomy component→concept ∪ panel) was embedded with no cross-proposer dedup — so genuine near-duplicates slipped through (a figure box ≈ an anatomy component on a solo bust; overlapping booru head classes on one head), embedding the same region twice and burning a slot against max_regions. Add detectors.dedupe_crops(): a greedy, high-IoU (default 0.85), kind-aware pass over the pending (crop, template) list right before embed_batch — drop boxes that overlap ≥ iou within the same kind, keep the highest score. The high threshold is deliberate: it collapses only true near-identical boxes while preserving intentional nested crops across scopes (a whole figure vs a small head component sit well below it) and distinct kinds (concept vs panel). Env-tunable DEDUPE_IOU (≥1.0 disables). Runs on CPU before the GPU work, so it cuts both embed cost and region count. Temporal (cross-frame) dedup deferred. Build marker 2026-07-01.2. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ttrj5P7upUTueSfoJcxEqa
This commit is contained in:
@@ -36,6 +36,8 @@ class Config:
|
||||
max_panels: int # cap panel crops per frame
|
||||
max_figures: int # cap figure boxes per frame (each = a CCIP call + crop)
|
||||
max_regions: int # hard cap on total regions per JOB (submit-size backstop)
|
||||
dedupe_iou: float # crops overlapping >= this (same kind) are near-dupes,
|
||||
# dropped before the embed; >=1.0 disables it
|
||||
|
||||
@classmethod
|
||||
def from_env(cls) -> Config:
|
||||
@@ -62,4 +64,5 @@ class Config:
|
||||
max_panels=int(os.environ.get("MAX_PANELS", "8")),
|
||||
max_figures=int(os.environ.get("MAX_FIGURES", "8")),
|
||||
max_regions=int(os.environ.get("MAX_REGIONS", "128")),
|
||||
dedupe_iou=float(os.environ.get("DEDUPE_IOU", "0.85")),
|
||||
)
|
||||
|
||||
Reference in New Issue
Block a user