Abstract
Convolutional neural networks (CNNs) have rapidly become the state-of-the-art models for image classification applications. They usually require large groundtruthed datasets for training. Here, we address object identification and recognition in the wild for infrared (IR) imaging in defense applications, where no such large-scale dataset is available. With a focus on robustness issues, especially viewpoint invariance, we introduce a compact and fully convolutional CNN architecture with global average pooling. We show that this model trained from realistic simulation datasets reaches a state-of-the-art performance compared with other CNNs with no data augmentation and fine-tuning steps. We also demonstrate a significant improvement in the robustness to viewpoint changes with respect to an operational support vector machine (SVM)-based scheme.
Highlights
As shown on reference datasets such as ImageNet [1], convolutional neural networks (CNNs) have become the state-of-the-art approaches for object classification in images
We focus on modular solutions and assume that we are provided with a target detection algorithm, which extracts image patches for a recognition and identification stage
Benchmarking experiments on real IR images patches demonstrate the relevance of the cfCNN with respect to state-of-the-art CNN architectures and an approach based on support vector machines (SVM)
Summary
As shown on reference datasets such as ImageNet [1], convolutional neural networks (CNNs) have become the state-of-the-art approaches for object classification in images. Since a high robustness is of key importance in defense applications, modular solutions that would be easier to understand and evaluate may be preferred, for instance, with distinct modules for detection and classification Based on this observation, we focus on modular solutions and assume that we are provided with a target detection algorithm, which extracts image patches for a recognition and identification stage. Benchmarking experiments on real IR images patches demonstrate the relevance of the cfCNN with respect to state-of-the-art CNN architectures and an approach based on support vector machines (SVM) They stress the importance of realistic synthetic dataset. Given the targeted operational contexts, we study the robustness of our cfCNN against possible perturbations introduced by the detection stage To simulate such behaviour, we train our network on images with centred targets and test it on translated or scaled inputs.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.