# PACGNet: Pyramidal Adaptive Cross-Gating Network for Multimodal Object Detection in Aerial Imagery

arXiv: https://arxiv.org/abs/2512.18291 · HTML: https://mlsanity.com/paper/2512.18291/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 41 references that are in the map.
- First posted on arXiv 20 Dec 2025.

## Builds on (most cited first, 20 of 20)

- [CBAM: Convolutional Block Attention Module](https://mlsanity.com/paper/1807.06521/) (2018)
- [Path Aggregation Network for Instance Segmentation](https://mlsanity.com/paper/1803.01534/) (2018)
- [ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks](https://mlsanity.com/paper/1910.03151/) (2019)
- [Oriented R-CNN for Object Detection](https://mlsanity.com/paper/2108.05699/) (2021)
- [Align Deep Features for Oriented Object Detection](https://mlsanity.com/paper/2008.09397/) (2020)
- [Drone-based RGB-Infrared Cross-Modality Vehicle Detection via Uncertainty-Aware Learning](https://mlsanity.com/paper/2003.02437/) (2020)
- [Cross-Modality Fusion Transformer for Multispectral Object Detection](https://mlsanity.com/paper/2111.00273/) (2021)
- [Multispectral Deep Neural Networks for Pedestrian Detection](https://mlsanity.com/paper/1611.02644/) (2016)
- [ICAFusion: Iterative Cross-Attention Guided Feature Fusion for Multispectral Object Detection](https://mlsanity.com/paper/2308.07504/) (2023)
- [Foreground-Aware Relation Network for Geospatial Object Segmentation in High Spatial Resolution Remote Sensing Imagery](https://mlsanity.com/paper/2011.09766/) (2020)
- [Fusion-Mamba for Cross-modality Object Detection](https://mlsanity.com/paper/2404.09146/) (2024)
- [C2Former: Calibrated and Complementary Transformer for RGB-Infrared Object Detection](https://mlsanity.com/paper/2306.16175/) (2023)
- [Multispectral Pedestrian Detection via Simultaneous Detection and Segmentation](https://mlsanity.com/paper/1808.04818/) (2018)
- [Cross-Modality Attentive Feature Fusion for Object Detection in Multispectral Remote Sensing Imagery](https://mlsanity.com/paper/2112.02991/) (2021)
- [Wise-IoU: Bounding Box Regression Loss with Dynamic Focusing Mechanism](https://mlsanity.com/paper/2301.10051/) (2023)
- [UAV-DETR: Efficient End-to-End Object Detection for Unmanned Aerial Vehicle Imagery](https://mlsanity.com/paper/2501.01855/) (2025)
- [Guided Hybrid Quantization for Object detection in Multimodal Remote Sensing Imagery via One-to-one Self-teaching](https://mlsanity.com/paper/2301.00131/) (2022)
- [Anchor-free Small-scale Multispectral Pedestrian Detection](https://mlsanity.com/paper/2008.08418/) (2020)
- [Nighttime Pedestrian Detection Based on Fore-Background Contrast Learning](https://mlsanity.com/paper/2408.03030/) (2024)
- [Selective Structured State Space for Multispectral-fused Small Target Detection](https://mlsanity.com/paper/2505.14043/) (2025)

## Abstract (arXiv)

Object detection in aerial imagery is challenging because targets are often small, backgrounds are cluttered, and imaging conditions vary substantially. Multimodal sensing can improve robustness by providing complementary visual and spectral information, but effective fusion must balance cross-modal interaction with computational cost. This paper presents PACGNet, a lightweight backbone-centric framework for multimodal aerial object detection. Its Symmetrical Cross-Gating (SCG) module performs bidirectional gated interaction between the two modality streams at the same feature scale while retaining residual pathways for the original representations. The Pyramidal Feature-aware Multimodal Gating (PFMG) module subsequently forms a single fused feature pyramid by using the preceding higher-resolution level as a spatial prior for fusion at the next level. PACGNet achieves 82.2% mAP50 on the held-out DroneVehicle test set and 82.1% mAP50 on VEDAI with only 5.2 M parameters and 13.2 GFLOPs. These results indicate a favorable accuracy-efficiency trade-off for lightweight multimodal aerial object detection. Code is available at https://github.com/Alianess/PACGNet.