Scalable Model-Assisted Multi-Target Estimation in Large Image Collections

2026-07-20

AI Summary

This work addresses the lack of statistical guarantees in existing computer vision models when estimating population-level quantities—such as class counts or proportions—in large-scale images. The authors propose a scalable estimation framework that integrates deep model predictions with limited human annotations, systematically adapting survey sampling techniques—including importance sampling, uniform sampling, control variates, and subset ratio estimation—to multi-object visual estimation for the first time. The resulting method yields unbiased, variance-controlled population estimates across diverse annotation budgets and numbers of target classes. Experiments on five detection and segmentation datasets spanning 7 to 80 categories demonstrate the framework’s effectiveness, significantly improving estimation accuracy under limited labeling resources and enabling robust, scientifically reliable measurements.

Abstract

Computer vision models are increasingly used as measurement tools to estimate population-level quantities from large image collections, but prediction errors introduce bias and the resulting estimates lack statistical guarantees required in scientific applications. Prior work uses a Monte Carlo framework to combine model predictions with ground-truth annotations by sampling some images for humans to label and is able to provide unbiased estimates with controllable accuracy, but primarily addresses single-scalar estimation. We study the more general problem of multi-target estimation, where many quantities (e.g., class counts or proportions) must be estimated simultaneously, and adapt sampling and estimation strategies from survey sampling to this setting. Evaluations on five detection and segmentation datasets with 7-80 classes show that importance sampling excels with moderate annotation budgets or fewer targets, whereas uniform sampling with control variates is superior when estimating many targets or operating with minimal labels. Additionally, a subset-based ratio estimator remains highly competitive across all regimes. Ultimately, our framework effectively combines biased model predictions and limited human labels into rigorous scientific measurements.

Problem

• multi-target estimation
• model bias
• statistical guarantees
• large image collections
• population-level quantities

Innovation

• multi-target estimation
• model-assisted estimation
• survey sampling
• importance sampling
• control variates