We are pleased to announce that our paper “From Object Difficulty to Image Scoring: A Strategy for Active Learning in Object Detection” has been accepted for publication in the Knowledge-Based Systems journal.

• About Knowledge-Based Systems

Authors:
Duc Tai Phan, Nhut Minh Nguyen, Khang Phuc Nguyen, Phuong-Nam Tran, Nhat Truong Pham, Linh Le, Choong Seon Hong, and Duc Ngoc Minh Dang

Abstract:
This work addresses the high annotation cost in object detection by proposing Feature Difficulty-based Active Learning (FDAL), a novel framework that selects the most informative samples for labeling by analyzing instability in the feature space. Unlike traditional approaches that rely mainly on output-level uncertainty, FDAL systematically interpolates object features toward class anchors in the latent space to reveal hidden characteristics that the model has not yet learned. By measuring prediction inconsistencies under these controlled perturbations, the framework captures both classification and localization difficulty in a unified metric.

Extensive experiments across four benchmark datasets—PASCAL VOC, KITTI, Cityscapes, and MS COCO—demonstrate that FDAL consistently outperforms state-of-the-art active learning strategies while significantly reducing labeling costs. The proposed method provides an efficient and scalable approach for improving object detection models in scenarios with limited annotated data.

We congratulate all authors on this achievement.

Share this article: