Component-Based OOD Detection decomposes inputs into functional components via Component Shift Score and Compositional Consistency Score, improving coarse- and fine-grained out-of-distribution detection without training.
M²E-UAV introduces the first onboard event-based benchmark for motion-on-motion tiny UAV detection, showing current methods fail under dense ego-motion and sparse targets.
Flow Mismatching detects anomalies via velocity discrepancies between normal flow dynamics and geometric paths to test images, yielding pixel heatmaps and image scores without test-time optimization and achieving state-of-the-art results.
Re-annotation reveals object detection benchmarks miss up to 60% of objects due to incomplete labels, and current detectors remain misaligned with human perception.
FDEP integrates frozen visual foundation model representations into infrared small target detection via semantic alignment fusion and implicit self-distillation, achieving state-of-the-art accuracy with no inference overhead.
DAD treats graphic design detection as ordered compositional deconstruction with amodal bounding boxes and element-level reinforcement learning, achieving human-level amodal detection and outperforming baselines across nine benchmarks.
GATE-AD employs graph attention networks with masked reconstruction to detect industrial anomalies from few normal samples, achieving state-of-the-art accuracy with faster inference across benchmarks.
FullTilt detects open-set 3D macromolecules directly from 2D tilt-series via a tilt-series encoder, accelerating inference orders of magnitude while achieving state-of-the-art zero-shot results.