Skip to content
Conference

CVDI-YOLO26: context-verified detail injection for small-object detection in high-altitude UAV infrared imagery

Oct 2026 · International Conference on Image Processing, Object Detection and Tracking · Vol 14506, pp. 145061M - 145061M-7 · 0 citations · 12 references
Engineering

TL;DR

Context-Verified Detail Injection (CVDI), which uses P2 only as a detail source and extracts a low-pass residual from the stride-4 feature, maps it to P3 through anti-aliased reduction, forms a context-conditioned acceptance map from P3 semantics and annular contrast, and injects detail through a zero-initialized residual.

Abstract

Small targets in high-altitude unmanned-aerial-vehicle (UAV) infrared imagery contain limited texture and are easily confused with thermal clutter. YOLO26n performs lightweight end-to-end detection on P3–P5 features, but repeated downsampling can suppress the local structure of tiny objects. Adding a P2 prediction head preserves detail at the cost of a denser grid and more background candidates. This paper presents Context-Verified Detail Injection (CVDI), which uses P2 only as a detail source. CVDI extracts a low-pass residual from the stride-4 feature, maps it to P3 through anti-aliased reduction, forms a context-conditioned acceptance map from P3 semantics and annular contrast, and injects detail through a zero-initialized residual. On the four-class HIT-UAV v1.2.1 split, the representative CVDI-YOLO26n result improves validation/test mAP50–95 from 0.62580/0.55462 to 0.63320/0.56240 and validation Small R from 0.85690 to 0.86650. Against a direct YOLO26n-P2 comparator, CVDI retains similar Small R while reducing GFLOPs. The module adds 0.3% parameters, 2.7% GFLOPs, and 3.9% fused FP16 batch-1 model-forward latency relative to YOLO26n. Three paired seeds report the same trend on both validation and test splits.

View source

Similar papers

Open access Sep 2026

MF-YOLO: High-Resolution Weighted Feature Fusion with Global-Local Aggregation for Small-Object Detection in UAV Imagery

MF-YOLO improves small-object detection at the cost of substantially greater computation, and introduces a stride-4 P2 prediction scale, applies Global-to-Local Spatial Aggregation independently to C2–C5 before neck fusion, and connects P2–P5 through a 64-channel Bidirectional Feature Pyramid Network (BiFPN).

Yu-Fei Wang, Ji-Ming Yang, Fang Liu et al. · 0 citations
Open access Sep 2026

RAD-YOLO: a lightweight edge-deployed visual measurement method for small-object detection in UAV imagery

This work proposes RAD-YOLO, a YOLOv8s-based small-object detector, and develops an edge-deployable variant named RAD-YOLO-Slim, which provides a practical balance of accuracy, speed and energy efficiency on the RK3576 platform.

Shuai-Jie Nie, Jia-Jian Yang, Xin He et al. · 0 citations
Open access Oct 2026

EHL-YOLO: hierarchical multi-scale feature extraction and efficient cross-level fusion for UAV object detection

EHL-YOLO, a task-oriented structural optimization of YOLOv10l that coordinates shallow feature preservation, hierarchical cross-scale interaction, and localization-quality-aware prediction, demonstrates an improved accuracy–complexity balance for visible-light UAV detection under the reported protocol, while the HIT-UA...

Shi He, Qiong-Li He, Yu-Hao Yan · 0 citations
Open access Oct 2026

SA-YOLO: Lightweight Adaptive Small-UAV Detection in Complex Urban Environments

Compared with the accuracy-leading detectors, SA-YOLO delivers comparable accuracy in a more compact model, achieving a favorable accuracy–model-size tradeoff at a real-time inference speed.

Xiao-Tian Zhang, Long-Miao Chen, Chao Wang et al. · 0 citations
Open access 2026

UAV-YOLO: A Resolution-Guided Feature Fusion and Attention Network for Small Object Detection in Aerial Imagery

Small and densely distributed objects in unmanned aerial vehicle (UAV) imagery are difficult to detect because deep feature extraction progressively removes the spatial evidence needed to separate and localize neighboring targets. This paper presents UAV-YOLO, a resolution-guided multi-scale detector that couples spati...

Jun-Hao Xu, Zeng-Hui Li · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.