Class competition reveals confidence in crowded object counting
How Many, but How Sure? Class Competition as a Confidence Proxy in Multi-Class Density-Based Counting
Machine Learning
Summary
Counting objects in crowded scenes is tricky when multiple types overlap, and it’s hard to know how sure the computer is about each count. The authors came up with a way to measure how much different object types compete locally in the prediction, which tells us how confident we can be without changing the existing model. They tested this on a known dataset and showed that adjusting counts based on this competition improved accuracy. This means the competition between predicted classes carries useful clues about how reliable the object counts are.
What this means in practice
- •For urban planners: Improve reliability of automated crowd counting tools by using class competition to highlight uncertain counts in public spaces.
- •For surveillance system engineers: Enhance multi-class object detection systems by using class competition as a confidence measure without retraining models.
Tested on one dataset.
Authors
Villanelle O'Reilly, Georgios Leontidis, Marc Hanheide
Abstract
Multi-class density estimators can count objects in crowded scenes, but their class-wise outputs offer no indication of how reliably density mass has been allocated. We introduce a training-free confidence proxy based on local competition between the strongest class channels. The resulting prominence measure requires only the predicted density maps, with no retraining or access to model internals. We test whether residual competition is diagnostic of cross-class leakage by attenuating contested density at inference time. On DSACA predictions for the Hicks et al. dataset, this reduces overall MAE from 0.920 to 0.846, suggesting that class competition contains useful information about count reliability.