What question did this study set out to answer?

The research aims to improve face detection accuracy under occlusion, for small-scale faces, and in imbalanced samples.

April 12, 2026Applied Sciences0 citationsOpen Access

Towards Accurate Face Detection Under Occlusion, Class Imbalance and Small-Scale Challenges

LLLinrunjia LiuXidian University DLDayong LiXidian University SWSha WuXidian University

Key Points

The research aims to improve face detection accuracy under occlusion, for small-scale faces, and in imbalanced samples.
Developed a YOLOv7-based detector incorporating an optimized MPConv module for feature extraction.
Implemented a CFPM to increase sensitivity to occluded faces.
Utilized DyHead block in IDetect to reduce feature loss from sample imbalance.
Applied an SW-SCE loss function with a dual-input network for better small face detection.
Improved detection performance by 1.2% for easy faces, 1.8% for medium faces, and 3% for hard faces compared to the baseline.
Enhanced capability to detect faces in dense environments with significant occlusion and small-scale targets.

Abstract

To address face occlusion, low detection rates of small-scale faces, and sample imbalance in dense visual scenarios, we propose a YOLOv7-based detector with four key improvements: (1) an optimized MPConv module to enhance feature extraction; (2) a novel CFPM to boost sensitivity to occluded samples; (3) an integration of the DyHead block in IDetect to mitigate feature loss from sample imbalance; (4) an SW-SCE loss function with a dual-input network to better detect small faces. Experiments on the WiderFace dataset show that our method improves detection performance by 1.2%, 1.8%, and 3% on the easy, medium, and hard subsets over the baseline. These gains strengthen face detection in dense, challenging environments with heavy occlusion and small-scale targets.

Perguntar à IA

Bookmark

View Full Paper