G-MAD: A Game-Based Data Generation Framework for Multi-View RGB-T Aerial Object Detection

Yechan Kim1, JongHyun Park2, Dongho Yoon2, Namhoon Jung1, Moongu Jeon2
1LIG Defense&Aerospace (LIG D&A), 2GIST
ACM Multimedia (MM) 2026
*Correspondence to yechan.kim26@ligdna.com or yechan0802@gmail.com
Paper Code Data
Teaser image

The overall architecture of our G-MAD framework.

Abstract

This work introduces G-MAD, an open-source framework that uses Arma3 to generate synchronized multi-view RGB-T data for aerial object detection. G-MAD addresses key limitations of real-world aerial dataset construction, including limited viewpoint control, imperfect RGB-T alignment and high annotation cost. The framework supports structured scenario specification, controllable multi-view camera placement, simultaneous visible/thermal capture, and automatic bounding box annotation using engine-level geometric metadata. These capabilities enable controlled studies of viewpoint variation, multi-modal fusion, and synthetic-to-real transfer in aerial object detection. Besides, using G-MAD, we construct and release [AMOD 👆], a new large-scale multi-view aerial RGB-T object detection benchmark.

Comparison of existing Arma3-based synthetic data generation frameworks

[1] [2] G-MAD (Proposed)
Scene configuration and sampling
Multi-map support × × ○
Max objects per image ≤ 1 ≤ 1 ≥ 1
Region of interest selection Manual Manual Automatic (Constraint-aware)
Camera viewpoint sampling Predefined orbit Unconstrained random User-defined sampling model
Weather variation Fog, Sun, Rain Fog, Sun, Rain Fog, Overcast, Sun, Rain, Snow
Occlusion check between camera and object × ○ ○
Annotation and output
Annotation method Post-hoc image differencing Post-hoc image differencing Engine-native ground truth
Bounding box support × HBB HBB, OBB
Resolution (Image size) Fixed Fixed Configurable
Resolution scaling (supersampling) × × ○
Execution and automation
MATLAB dependency ○ × (Only Python) × (Only Python)
Execution scheme GUI-dependent control GUI-dependent control GUI-independent launch
Temporary file cleanup Manual Manual Automatic
Workstation availability during generation Unavailable Unavailable Available

[1] https://github.com/cloftus96/Synthetic-Data-Generation

[2] https://github.com/ttsiapras/Arma3DatasetGen

Potential applications of G-MAD

01

Multi-view Learning Benchmark

Benchmark and advance multi-view learning methods with diverse aerial viewpoints and structured scene variations.

02

RGB-T Fusion and Cross-modal Learning

Develop and evaluate RGB-T fusion pipelines and cross-modal representation learning approaches in a controlled setting.

03

Pretraining for Real-world Transfer

Use G-MAD as an effective pretraining source generator for transfer to real-world aerial domains and downstream perception tasks.

Tip: Arma3?

Teaser image

Arma3 is a military tactical shooter video game developed and published by Bohemia Interactive. It’s well-known in the gaming and simulation community for its realism, large-scale combat environments, and modding capabilities. For more information, please visit the official website of Arma3: https://arma3.com/.

BibTeX

@inproceedings{G-MAD2026,
  title={G-MAD: A Game-Based Data Generation Framework for Multi-View RGB-T Aerial Object Detection},
  author={Kim, Yechan and Park, JongHyun and Yoon, Dongho and Jung, Namhoon and Jeon, Moongu},
  year={2026},
  booktitle={ACM International Conference on Multimedia}
}