What question did this study set out to answer?

The aim is to decrease the costs of data annotation for object detection while maintaining model accuracy.

Adaptive Active Learning with Dynamic Uncertainty-Diversity Balancing for Object Detection

View Full Paper

Ask AI

Bookmark

Authors

This framework reduces annotation costs in object detection, implying greater efficiency for industrial applications.

The aim is to decrease the costs of data annotation for object detection while maintaining model accuracy.
Developed an adaptive active learning framework tailored for object detection.
Utilized confidence scores to estimate uncertainty in the data.
Quantified diversity based on class distribution in unlabeled images.
Dynamically adjusted the balance between uncertainty and diversity during training.
Conducted experiments on UAV and industrial datasets.
Achieved a performance improvement of 1.1% on the UAV dataset.
Showed approximately 10% enhancement on the industrial dataset under the same annotation budget.
Demonstrated the framework's effectiveness in real-world industrial scenarios.

Auh et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark