Camouflage Recognition in Military Operations using
Deep Learning
Gaddipati Vedasri Lakshmi Annadi Spoorthi Reddy Kosaraju Joshitha
22WH1A05D5 22WH1A05F2 22WH1A05G4
Under the Guidance of
Dr. Venkatesh B
Associate Professor,
Dept of C.S.E,
BVRIT HYDERABAD College of Engineering for Women.
November 17, 2025
Table of Content
1 Abstract
2 Introduction
3 Literature Survey
4 Technology
5 Datasets
6 Expected Outcome
7 References
Batch No C_11 Department of CSE November 17, 2025 2 / 12
Abstract
Camouflaged object detection is a growing area of interest in defense technology due to the
increasing sophistication of concealment methods used in military environments. Modern
surveillance systems require intelligent models capable of recognizing objects that blend
seamlessly into backgrounds under varying illumination, terrain, and environmental con-
ditions. This project presents a deep learning–based framework that combines YOLO for
real-time object detection and SINet for pixel-level segmentation of camouflaged targets.
The system incorporates attention mechanisms and multi-modal imaging (RGB, infrared,
thermal) to enhance feature extraction and robustness. Experiments are conducted using
camouflage-specific datasets such as MHCD2022, CAMO, and Military Assets Detection.
Using PyTorch and OpenCV, the proposed approach provides an efficient and deployable
solution for defense-oriented detection applications.
Batch No C_11 Department of CSE November 17, 2025 3 / 12
Introduction
Modern defense operations demand advanced surveillance systems capable of identifying
threats even when they are intentionally concealed within natural or artificial environ-
ments. Conventional vision-based methods fall short due to background complexity, vari-
able lighting, and the subtle textures of camouflaged targets. To address these limitations,
this project explores a deep learning–driven approach for both detection and segmentation
of hidden objects. Using YOLO for rapid object localization and SINet for detailed seg-
mentation, the framework integrates attention modules and multi-modal imaging sources
(RGB, infrared, thermal) to capture fine-grained visual cues. This combined strategy en-
hances reliability and precision, making it suitable for real-world military surveillance and
reconnaissance applications.
Batch No C_11 Department of CSE November 17, 2025 4 / 12
Literature Survey
[Link]. Title Methods Evaluation Pros Cons
Metrics
1 Chutke et al. (2025) Hybrid deep Precision Useful for defence Not real-time
AI-Enhanced Image learning F1 Score surveillance Not camouflage-
Processing for De- Preprocessing Strong perfor- specific
fence Detection & detection mance in noise
pipeline
2 Cools et al. (2025) Visual & thermal Human Useful for camou- No deep learning
Evaluation Tech- evaluation perception flage design usage
niques for Military Standardized metrics Strong evaluation Not a detection
Camouflage testing frame- Pattern er- guidelines model
work ror analysis
3 Zhang et al. (2025) Vision-language Precision Excellent in clut- Requires text input
Referring Cam- fusion Recall tered scenes High computation
ouflaged Object Referring expres- Introduces new cost
Detection sion grounding dataset
Batch No C_11 Department of CSE November 17, 2025 5 / 12
Literature Survey
[Link]. Title Methods Evaluation Pros Cons
Metrics
4 Li et al. (2024) DETR trans- mAP Strong global rea- Slow inference
MilDetr Trans- former soning Needs large
former Multi-head at- Good for complex datasets
tention scenes
5 Song et al. (2024) Multi-view fu- S-measure Works on salient & Requires multi-
MVGN Network sion MAE camouflaged view data
Shared encoder High generalization High training cost
6 Truong & Tran Neural style mAP Increases dataset Artifacts possible
(2024) transfer diversity Sensitive to style
Style Transfer Aug- Texture mixing Boosts accuracy selection
mentation
7 Kamble & Rajara- CNN-based ar- Accuracy Lightweight Weak in complex
jeswari (2024) chitecture Easy to train scenes
CNN Camouflage Poor generalization
Detection
Batch No C_11 Department of CSE November 17, 2025 6 / 12
Literature Survey
[Link]. Title Methods Evaluation Pros Cons
Metrics
8 Liu et al. (2023) Semi-supervised MAE Works with limited Sensitive to low
Semi-supervised SINet labels resolution
SINet Search- Effective for hu- Needs pseudo-label
identification man camouflage refinement
module
9 Liu & Di (2023) MHNet deep MAE Excellent for mili- Heavy model
MHNet model Accuracy tary textures Dataset limited to
MH camouflage Strong feature ex- military domain
dataset traction
10 Wei et al. (2022) Enhanced mAP Real-time detec- Fails in extreme
Improved YOLOv5 YOLOv5 FPS tion camouflage
Feature fusion Good object accu- Limited robustness
improvements racy
Batch No C_11 Department of CSE November 17, 2025 7 / 12
Technology
Packages
Python
PyTorch – Deep learning framework
Torchvision – Pretrained models & image transforms
YOLO (You Only Look Once) – Object detection
SINet (Selective Information Network) – Camouflage object segmentation
OpenCV – Image processing
CNN (Convolutional Neural Network) - Feature extraction
Batch No C_11 Department of CSE November 17, 2025 8 / 12
Datasets
Table 1: Key Datasets for Camouflage Detection
Dataset Link
MHCD2022 [Link]
mhcd2022
CAMO Dataset [Link]
ivanomelchenkoim11/camo-dataset
Military Assets Detection [Link]
military-assets-detection-yolov8/input
Batch No C_11 Department of CSE November 17, 2025 9 / 12
Expected Outcome
Figure 1: Military camouflage
Batch No C_11 Department of CSE November 17, 2025 10 / 12
References
1) Chutke, Sachin, V. V. Kumar, M. Saravanakumar, P. K. Lendale, M. Akku, and K. S.
Krishna. "AI-Enhanced Image Processing for Target Detection and Threat Recognition
in Defence Applications." In 2025 6th International Conference on Intelligent Commu-
nication Technologies and Virtual Mobile Networks (ICICV), pp. 763–768. IEEE, 2025.
2) Cools, Kasper, Clara Maathuis, Geert De Cubber, Marijke VandewaL, and Nikos Deli-
giannis. "Evaluation Techniques for Modern Military Camouflage." In 2025 International
Conference on Military Technologies (ICMT), pp. 1–6. IEEE, 2025.
3) Zhang, Xuying, Bowen Yin, Zheng Lin, Qibin Hou, Deng-Ping Fan, and Ming-Ming
Cheng. "Referring camouflaged object detection." IEEE Transactions on Pattern Analy-
sis and Machine Intelligence (2025).
4) Li, Bo, Ren Zhou, Liang Yang, Qian Wang, and Huan Chen. "MilDetr: Detection Trans-
former for Military Camouflaged Target Detection." IEEE Access 12 (2024): 26163–
26174.
5) Song, Xiaogang, Pengfei Zhang, Xiaofeng Lu, Xinhong Hei, and Rongrong Liu. "A uni-
versal multi-view guided network for salient object and camouflaged object detection."
IEEE Transactions on Circuits and Systems for Video Technology 34, no. 11 (2024):
11184–11197.
6) Truong, Thi Thu Hang, and Trung Kien Tran. "A style transfer-based augmentation ap-
proach for detecting military camouflaged object." JMST’s Section on Computer Science
and Control Engineering CSCE8 (2024): 44–54.
Batch No C_11 Department of CSE November 17, 2025 11 / 12
References
7) Kamble, R., and P. Rajarajeswari. "Revealing Hidden Patterns: A Deep Learning Ap-
proach to Camouflage Detection." International Journal of Computational Methods and
Experimental Measurements 12, no. 1 (2024): 97–105.
8) Liu, Y., C. Q. Wang, and Y. J. Zhou. "Camouflaged people detection based on a semi-
supervised search identification network." Defence Technology 21 (2023): 176–183.
9) Liu, Ming, and Xue Di. "Extraordinary MHNet: Military high-level camouflage object
detection network and dataset." Neurocomputing 549 (2023): 126466.
10) Wei, Z., Z. Qikai, L. Ruizhi, and N. Fu. "Research on camouflage target detection method
based on improved YOLOv5." Journal of Physics: Conference Series 2284, no. 1 (2022).
Batch No C_11 Department of CSE November 17, 2025 12 / 12