Work

Deep Learning System for Hierarchical Identification and Classification of Military Vehicles in Images

Year: 2026

Type: book-chapter

Abstract: This paper addresses the problem of automated identification and classification of military equipment from images using deep learning methods. A dataset preparation approach is proposed, incorporating hierarchical class organization and grouping of tank modifications by base types. A client-server system based on React, Python, and PyTorch is developed to enable real-time image processing. A comparative analysis of ResNet50 and EfficientNet-B0 is conducted. Both models demonstrate high accuracy at the base classification level, while performance decreases significantly for fine-grained classification due to strong visual similarity between modifications. ResNet50 achieves higher accuracy in this scenario. The use of transfer learning, Focal Loss, and advanced augmentation techniques improves performance and reliability. The results confirm the effectiveness of the proposed approach for automated analysis of military imagery under limited and imbalanced data conditions.

Source: Frontiers in artificial intelligence and applications

DOI: 10.3233/faia260659

Authors: Renjie Cheng, Dmytro Uhryn, Yurii Ushenko, Artem Karachevtsev, Oleh Pastukh, Oksana Yatsko

Institutions: Hubei University of Technology Engineering and Technology College, Yuriy Fedkovych Chernivtsi National University, Ternopil Ivan Pului National Technical University

Language: English

Topic: Advanced Neural Network Applications

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: hybrid

OpenAlex ID: W7204544010

FWCI:

Cites:

Cited by:

Related to: 10