Cascade R-CNN: stage-specific detection at increasing overlap thresholds

What changed

Trained a sequence of detector heads at increasing IoU quality thresholds, using progressively refined proposals to address sample scarcity and train–inference quality mismatch.

Scientific scope

Trained a sequence of detector heads at increasing IoU quality thresholds, using progressively refined proposals to address sample scarcity and train–inference quality mismatch. Not merely repeated application of one fixed regressor: stages are trained for different proposal distributions/thresholds. Extra stages add computation. Empirical sample-quality and accuracy improvements are not a guarantee of monotone localization for every object.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Not merely repeated application of one fixed regressor: stages are trained for different proposal distributions/thresholds. Extra stages add computation. Empirical sample-quality and accuracy improvements are not a guarantee of monotone localization for every object.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Cascade R-CNN: Delving into High Quality Object Detection

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction