2026
Journal article
Open Access
A Survey on SAR ship classification using deep learning
Ch Muhammad Awais, Marco Reggiannini, Davide Moroni, Emanuele Salerno
Deep learning (DL) has become a central approach for ship classification using synthetic aperture radar (SAR) imagery. This survey reviews 74 representative studies selected from 187 publications, categorizing them into a taxonomy with four main dimensions: (i) DL architectures, (ii) datasets, (iii) image augmentation, and (iv) learning techniques. We analyze how approaches such as handcrafted feature integration, data augmentation, fine-tuning, and transfer learning influence classification performance, and summarize the use of public benchmarks including OpenSARShip and FUSARShip. This survey highlights key challenges: limited data availability, class imbalance, lack of standardized metrics, and limited interpretability of DL models. Future research directions include the development of SAR-specific DL architectures, advanced augmentation and generative approaches, integration of handcrafted and deep features, interpretable DL, and stronger interdisciplinary collaboration. By addressing these challenges, DL-based SAR ship classification can achieve greater robustness, accuracy, and transparency, ultimately strengthening maritime surveillance and operational monitoring.Source: IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
DOI: 10.1109/jstars.2026.3695704
Metrics:
Ch Muhammad Awais, Marco Reggiannini, Davide Moroni, Emanuele Salerno
Deep learning (DL) has become a central approach for ship classification using synthetic aperture radar (SAR) imagery. This survey reviews 74 representative studies selected from 187 publications, categorizing them into a taxonomy with four main dimensions: (i) DL architectures, (ii) datasets, (iii) image augmentation, and (iv) learning techniques. We analyze how approaches such as handcrafted feature integration, data augmentation, fine-tuning, and transfer learning influence classification performance, and summarize the use of public benchmarks including OpenSARShip and FUSARShip. This survey highlights key challenges: limited data availability, class imbalance, lack of standardized metrics, and limited interpretability of DL models. Future research directions include the development of SAR-specific DL architectures, advanced augmentation and generative approaches, integration of handcrafted and deep features, interpretable DL, and stronger interdisciplinary collaboration. By addressing these challenges, DL-based SAR ship classification can achieve greater robustness, accuracy, and transparency, ultimately strengthening maritime surveillance and operational monitoring.Source: IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
DOI: 10.1109/jstars.2026.3695704
Metrics:
See at:
CNR IRIS 
| ieeexplore.ieee.org | CNR IRIS 
2026
Journal article
Open Access
A classification-aware super-resolution framework for ship targets in SAR imagery
Awais Ch Muhammad, Reggiannini Marco, Moroni Davide, Karakus Oktay
High-resolution imagery plays a critical role in improving the performance of visual recognition tasks such as classification, detection, and segmentation. In many domains, including remote sensing and surveillance, low-resolution images can limit the accuracy of automated analysis. To address this, super-resolution (SR) techniques have been widely adopted to attempt to reconstruct high-resolution images from low-resolution inputs. Related traditional approaches focus solely on enhancing image quality based on pixel-level metrics, leaving the relationship between super-resolved image fidelity and downstream classification performance largely underexplored. This raises a key question: can integrating classification objectives directly into the super-resolution process further improve classification accuracy? In this paper, we try to respond to this question by investigating the relationship between super-resolution and classification through the deployment of a specialised algorithmic strategy. We propose a novel methodology that increases the resolution of synthetic aperture radar imagery by optimising loss functions that account for both image quality and classification performance. Our approach improves image quality, as measured by scientifically ascertained image quality indicators, while also enhancing classification accuracy.Source: IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, vol. 19, pp. 6614-6622
DOI: 10.1109/jstars.2026.3655550
Metrics:
Awais Ch Muhammad, Reggiannini Marco, Moroni Davide, Karakus Oktay
High-resolution imagery plays a critical role in improving the performance of visual recognition tasks such as classification, detection, and segmentation. In many domains, including remote sensing and surveillance, low-resolution images can limit the accuracy of automated analysis. To address this, super-resolution (SR) techniques have been widely adopted to attempt to reconstruct high-resolution images from low-resolution inputs. Related traditional approaches focus solely on enhancing image quality based on pixel-level metrics, leaving the relationship between super-resolved image fidelity and downstream classification performance largely underexplored. This raises a key question: can integrating classification objectives directly into the super-resolution process further improve classification accuracy? In this paper, we try to respond to this question by investigating the relationship between super-resolution and classification through the deployment of a specialised algorithmic strategy. We propose a novel methodology that increases the resolution of synthetic aperture radar imagery by optimising loss functions that account for both image quality and classification performance. Our approach improves image quality, as measured by scientifically ascertained image quality indicators, while also enhancing classification accuracy.Source: IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, vol. 19, pp. 6614-6622
DOI: 10.1109/jstars.2026.3655550
Metrics:
See at:
CNR IRIS | ieeexplore.ieee.org | CNR IRIS | CNR IRIS
2025
Conference article
Open Access
Feature-space oversampling for addressing class imbalance in SAR ship classification
Awais Ch Muhammad, Reggiannini M., Moroni D., Karakuş O.
SAR ship classification faces the challenge of long-tailed datasets, which complicates the classification of underrepresented classes. Oversampling methods have proven effective in addressing class imbalance in optical data. In this paper, we evaluated the effect of oversampling in the feature space for SAR ship classification. We propose two novel algorithms inspired by the Major-to-minor (M2m) method M2mf, M2mu. The algorithms are tested on two public datasets, OpenSARShip (6 classes) and FuSARShip (9 classes), using three state-of-the-art models as feature extractors: ViT, VGG16, and ResNet50. Additionally, we also analyzed the impact of oversampling methods on different class sizes. The results demonstrated the effectiveness of our novel methods over the original M2m and baselines, with an average F1-score increase of 8.82% for FuSARShip and 4.44% for OpenSARShip.DOI: 10.1109/igarss55030.2025.11242334
DOI: 10.48550/arxiv.2508.06420
Metrics:
Awais Ch Muhammad, Reggiannini M., Moroni D., Karakuş O.
SAR ship classification faces the challenge of long-tailed datasets, which complicates the classification of underrepresented classes. Oversampling methods have proven effective in addressing class imbalance in optical data. In this paper, we evaluated the effect of oversampling in the feature space for SAR ship classification. We propose two novel algorithms inspired by the Major-to-minor (M2m) method M2mf, M2mu. The algorithms are tested on two public datasets, OpenSARShip (6 classes) and FuSARShip (9 classes), using three state-of-the-art models as feature extractors: ViT, VGG16, and ResNet50. Additionally, we also analyzed the impact of oversampling methods on different class sizes. The results demonstrated the effectiveness of our novel methods over the original M2m and baselines, with an average F1-score increase of 8.82% for FuSARShip and 4.44% for OpenSARShip.DOI: 10.1109/igarss55030.2025.11242334
DOI: 10.48550/arxiv.2508.06420
Metrics:
See at:
arXiv.org e-Print Archive | CNR IRIS | ieeexplore.ieee.org | doi.org | doi.org | CNR IRIS | CNR IRIS
2025
Other
Open Access
SI-Lab Annual Research Report 2024
Awais Ch Muhammad, Baiamonte A., Benassi A., Berti A., Bertini G., Buongiorno R, Bulotta D., Cafiso M., Carboni A., Carloni G., Caudai C., Colantonio S., Conti F., Daoudagh S., Del Corso G., Fusco G., Galesi G., Germanese D., Gravili S., Ignesti G., Kuruoglu E. E., Lazzini G., Leone G. R., Leporini B., Magrini M., Martinelli M., Omrani Ali Reza, Pachetti E., Papini O., Paradisi P., Pardini F., Pascali M. A., Pieri G., Reggiannini M., Righi M., Salerno E., Salvetti O., Scozzari A., Sebastiani L., Straface S., Tampucci M., Tarabella L., Tonazzini A., Moroni D.
The Signal & Images Laboratory (SI-Lab) is an interdisciplinary research group in computer vision, signal analysis, intelligent vision systems and multimedia data understanding. It is part of the Institute of Information Science and Technologies (ISTI) of the National Research Council of Italy (CNR). This report accounts for the research activities of the Signal and Images Laboratory of the Institute of Information Science and Technologies during the year 2024.DOI: 10.32079/isti-ar-2025/002
Metrics:
Awais Ch Muhammad, Baiamonte A., Benassi A., Berti A., Bertini G., Buongiorno R, Bulotta D., Cafiso M., Carboni A., Carloni G., Caudai C., Colantonio S., Conti F., Daoudagh S., Del Corso G., Fusco G., Galesi G., Germanese D., Gravili S., Ignesti G., Kuruoglu E. E., Lazzini G., Leone G. R., Leporini B., Magrini M., Martinelli M., Omrani Ali Reza, Pachetti E., Papini O., Paradisi P., Pardini F., Pascali M. A., Pieri G., Reggiannini M., Righi M., Salerno E., Salvetti O., Scozzari A., Sebastiani L., Straface S., Tampucci M., Tarabella L., Tonazzini A., Moroni D.
The Signal & Images Laboratory (SI-Lab) is an interdisciplinary research group in computer vision, signal analysis, intelligent vision systems and multimedia data understanding. It is part of the Institute of Information Science and Technologies (ISTI) of the National Research Council of Italy (CNR). This report accounts for the research activities of the Signal and Images Laboratory of the Institute of Information Science and Technologies during the year 2024.DOI: 10.32079/isti-ar-2025/002
Metrics:
2025
Conference article
Open Access
Deep learning–based detection of Nephrops norvegicus burrows
Papini O., Cecapolli E., Domenichetti F., Pieri G., Reggiannini M., Zacchetti L., Martinelli M.
This work presents a methodology to support the assessment of a benthic species of great commercial importance (Nephrops norvegicus) taking advantage of a combination of machine learning and computer vision methods. Up to the present, abundance indices based on the density of this species are evaluated by visual inspection of underwater imagery and through manual counting of the observed burrows. A novel approach is proposed, based on the integration in the processing pipeline of a supervised learning model in charge of detecting the burrows. The model is trained exploiting underwater videos that experts annotate by identifying the frames where burrows are present and specifying the related features. To pursue such a goal, the proposed automated procedure must cope with several environmental issues, such as high underwater turbidity, uneven illumination, heavy colour distortions, as well as complexities arising from the presence of ambiguous objects and morphological features that may affect the misclassification rate. The proposed method was developed on video material collected in a specific area, but has the potential to be applied throughout the species' distribution range. Preliminary results concerning the analysis of data captured in the central Adriatic Sea are presented and discussed.DOI: 10.1109/metrosea66681.2025.11245690
Project(s): NAUTILOS
Metrics:
Papini O., Cecapolli E., Domenichetti F., Pieri G., Reggiannini M., Zacchetti L., Martinelli M.
This work presents a methodology to support the assessment of a benthic species of great commercial importance (Nephrops norvegicus) taking advantage of a combination of machine learning and computer vision methods. Up to the present, abundance indices based on the density of this species are evaluated by visual inspection of underwater imagery and through manual counting of the observed burrows. A novel approach is proposed, based on the integration in the processing pipeline of a supervised learning model in charge of detecting the burrows. The model is trained exploiting underwater videos that experts annotate by identifying the frames where burrows are present and specifying the related features. To pursue such a goal, the proposed automated procedure must cope with several environmental issues, such as high underwater turbidity, uneven illumination, heavy colour distortions, as well as complexities arising from the presence of ambiguous objects and morphological features that may affect the misclassification rate. The proposed method was developed on video material collected in a specific area, but has the potential to be applied throughout the species' distribution range. Preliminary results concerning the analysis of data captured in the central Adriatic Sea are presented and discussed.DOI: 10.1109/metrosea66681.2025.11245690
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | ieeexplore.ieee.org | doi.org | CNR IRIS | CNR IRIS
2025
Conference article
Open Access
Image quality vs performance in super-resolution for SAR ship classification
Awais Ch Muhammad, Reggiannini M., Moroni D.
Synthetic Aperture Radar (SAR) images for ship classification often face the problem of low resolution. Techniques like super-resolution (SR) can help to enhance the images for better ship classification. In this paper, we compared traditional interpolation techniques (bilinear, bicubic, Lanczos, nearest-neighbor) with deep learning SR methods (EDSR, RCAN, CARN) at 2x and 4x resolutions to analyze their effect in terms of image quality and classification performance. The image quality was assessed using metrics like Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). The findings indicate that while 2x resolution images typically achieved higher image quality scores, the 4x images often performed equally well or better in classification tasks. We utilized two versions of VGG: SR techniques yielded similar scores with a simple VGG, whereas, in the multi-scale VGG (MSVGG), traditional interpolation methods outperformed deep learning methods. Experiments confirm that super-resolved images reach high scores in terms of classical image quality metrics. However, this does not always translate directly into improved performance in SAR ship classification. This highlights the need to select SR techniques by jointly evaluating image quality metrics and classification performance.DOI: 10.1109/iscas56072.2025.11043629
Project(s): National Biodiversity Future Center
Metrics:
Awais Ch Muhammad, Reggiannini M., Moroni D.
Synthetic Aperture Radar (SAR) images for ship classification often face the problem of low resolution. Techniques like super-resolution (SR) can help to enhance the images for better ship classification. In this paper, we compared traditional interpolation techniques (bilinear, bicubic, Lanczos, nearest-neighbor) with deep learning SR methods (EDSR, RCAN, CARN) at 2x and 4x resolutions to analyze their effect in terms of image quality and classification performance. The image quality was assessed using metrics like Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). The findings indicate that while 2x resolution images typically achieved higher image quality scores, the 4x images often performed equally well or better in classification tasks. We utilized two versions of VGG: SR techniques yielded similar scores with a simple VGG, whereas, in the multi-scale VGG (MSVGG), traditional interpolation methods outperformed deep learning methods. Experiments confirm that super-resolved images reach high scores in terms of classical image quality metrics. However, this does not always translate directly into improved performance in SAR ship classification. This highlights the need to select SR techniques by jointly evaluating image quality metrics and classification performance.DOI: 10.1109/iscas56072.2025.11043629
Project(s): National Biodiversity Future Center
Metrics:
See at:
CNR IRIS | ieeexplore.ieee.org | CNR IRIS | CNR IRIS
2025
Conference article
Open Access
A framework for imbalanced SAR ship classification: curriculum learning, weighted loss functions, and a novel evaluation metric
Awais Ch Muhammad, Reggiannini M., Moroni D.
Accurate ship classification is essential for maritime traffic monitoring applications but is significantly hindered by imbalanced datasets. In this paper, we propose a novel methodology that combines curriculum learning with weighted loss functions to address class imbalance in the FUSAR-Ship dataset, facilitating the accurate classification of its nine classes. Our method achieved notable improvements, including a 6.53% average increase in F1-scores compared to baseline models, and successfully identified all classes, including previously misclassified ones. To better evaluate model performance on long-tailed datasets, we introduce a novel evaluation metric that provides a more nuanced assessment of classification ability across underrepresented classes. While demonstrated on the FUSAR-Ship dataset, our approach and metric are broadly applicable to other imbalanced classification problems.DOI: 10.1109/wacvw65960.2025.00171
Project(s): National Biodiversity Future Center
Metrics:
Awais Ch Muhammad, Reggiannini M., Moroni D.
Accurate ship classification is essential for maritime traffic monitoring applications but is significantly hindered by imbalanced datasets. In this paper, we propose a novel methodology that combines curriculum learning with weighted loss functions to address class imbalance in the FUSAR-Ship dataset, facilitating the accurate classification of its nine classes. Our method achieved notable improvements, including a 6.53% average increase in F1-scores compared to baseline models, and successfully identified all classes, including previously misclassified ones. To better evaluate model performance on long-tailed datasets, we introduce a novel evaluation metric that provides a more nuanced assessment of classification ability across underrepresented classes. While demonstrated on the FUSAR-Ship dataset, our approach and metric are broadly applicable to other imbalanced classification problems.DOI: 10.1109/wacvw65960.2025.00171
Project(s): National Biodiversity Future Center
Metrics:
See at:
CNR IRIS | ieeexplore.ieee.org | CNR IRIS | CNR IRIS
2025
Conference article
Open Access
Machine learning approaches for automated detection of nephrops norvegicus burrows in underwater surveys
Papini O., Cecapolli E., Domenichetti F., Martinelli M., Pieri G., Reggiannini M., Zacchetti L.
This paper presents an analysis of computer vision methods designed to automate the detection, recognition, and classification of Nephrops norvegicus burrows in underwater videos. The proposed approach seeks to evaluate the accuracy, minimise human error, and standardise the existing manual video analysis process. By leveraging machine learning techniques, the system described in this paper autonomously processes video streams and identifies N. norvegicus burrow openings on the seabed. Additionally, this study investigates data augmentation algorithms to expand an annotated dataset and evaluates the performances of the first results under different configurations.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15616, pp. 285-294. Kolkata, India, 01-05/12/2024
DOI: 10.1007/978-3-031-87663-9_24
Project(s): NAUTILOS
Metrics:
Papini O., Cecapolli E., Domenichetti F., Martinelli M., Pieri G., Reggiannini M., Zacchetti L.
This paper presents an analysis of computer vision methods designed to automate the detection, recognition, and classification of Nephrops norvegicus burrows in underwater videos. The proposed approach seeks to evaluate the accuracy, minimise human error, and standardise the existing manual video analysis process. By leveraging machine learning techniques, the system described in this paper autonomously processes video streams and identifies N. norvegicus burrow openings on the seabed. Additionally, this study investigates data augmentation algorithms to expand an annotated dataset and evaluates the performances of the first results under different configurations.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15616, pp. 285-294. Kolkata, India, 01-05/12/2024
DOI: 10.1007/978-3-031-87663-9_24
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | link.springer.com | CNR IRIS | CNR IRIS
2025
Conference article
Open Access
Machine learning for the evaluation of the Nephrops norvegicus Population
Reggiannini M., Martinelli M., Papini O., Zacchetti L., Domenichetti F., Pieri G.
This paper introduces computer vision methods for detecting, recognising, and estimating Nephrops norvegicus (Norway lobster) burrow density via Underwater Television surveys. The current manual approach involves human operators visually assessing videos, which is prone to errors and subjectivity. Automated machine learning systems show promise in identifying and counting burrows, potentially standardising recognition and reducing operator errors. However, challenges exist in implementing computer vision techniques. An automated system aims to process video streams, detect seabed openings, extract visual features, and classify N. norvegicus burrows, significantly advancing the automation of underwater video reading. The primary processing presented in the paper lies in a boosting algorithm capable of extending the original annotated ground truth and assessing the improved performance of the extended data set with respect to the original one.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15509, pp. 282-295. Riva del Sole Resort & SPA - Castiglione della Pescaia, Toscana, 22-25/09/2024
DOI: 10.1007/978-3-031-82484-5_21
Project(s): NAUTILOS
Metrics:
Reggiannini M., Martinelli M., Papini O., Zacchetti L., Domenichetti F., Pieri G.
This paper introduces computer vision methods for detecting, recognising, and estimating Nephrops norvegicus (Norway lobster) burrow density via Underwater Television surveys. The current manual approach involves human operators visually assessing videos, which is prone to errors and subjectivity. Automated machine learning systems show promise in identifying and counting burrows, potentially standardising recognition and reducing operator errors. However, challenges exist in implementing computer vision techniques. An automated system aims to process video streams, detect seabed openings, extract visual features, and classify N. norvegicus burrows, significantly advancing the automation of underwater video reading. The primary processing presented in the paper lies in a boosting algorithm capable of extending the original annotated ground truth and assessing the improved performance of the extended data set with respect to the original one.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15509, pp. 282-295. Riva del Sole Resort & SPA - Castiglione della Pescaia, Toscana, 22-25/09/2024
DOI: 10.1007/978-3-031-82484-5_21
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | link.springer.com | doi.org | CNR IRIS | CNR IRIS
2025
Contribution to book
Open Access
Computer vision to support Nephrops norvegicus imagery annotation
Aguzzi J., Aristegui M., Burgos C., Chatzievangelou D., Doyle J., González Herraiz I., Fifas S., Firmin C., Jónasson J. P., Verisimo Amor P., Jonsson P., Martinelli M., Medvešek D., Pereira B., Reggiannini M., Schubert P., Vacherot J-. P., Valeiras J., Vila Y., Weetman A., Wieland K.
This document reports about the implementation of a computer vision procedure to estimate Nephrops norvegicus burrows density by analysing Underwater Television (UWTV) surveys. This activity, developed in cooperation with the ICES WGNEPS group, aims at providing an automatic system to support (i) the detection of the N. norvegicus openings, (ii) their grouping into systems (i.e. burrows) and (iii) the count of the distinct burrows. This could represent a relevant tool to simplify and optimise the stock assessment process.Source: ICES SCIENTIFIC REPORTS, pp. 23-28
DOI: 10.17895/ices.pub.28652012
Project(s): NAUTILOS
Metrics:
Aguzzi J., Aristegui M., Burgos C., Chatzievangelou D., Doyle J., González Herraiz I., Fifas S., Firmin C., Jónasson J. P., Verisimo Amor P., Jonsson P., Martinelli M., Medvešek D., Pereira B., Reggiannini M., Schubert P., Vacherot J-. P., Valeiras J., Vila Y., Weetman A., Wieland K.
This document reports about the implementation of a computer vision procedure to estimate Nephrops norvegicus burrows density by analysing Underwater Television (UWTV) surveys. This activity, developed in cooperation with the ICES WGNEPS group, aims at providing an automatic system to support (i) the detection of the N. norvegicus openings, (ii) their grouping into systems (i.e. burrows) and (iii) the count of the distinct burrows. This could represent a relevant tool to simplify and optimise the stock assessment process.Source: ICES SCIENTIFIC REPORTS, pp. 23-28
DOI: 10.17895/ices.pub.28652012
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | ices-library.figshare.com | CNR IRIS
2025
Journal article
Open Access
Advancing automated detection of Nephrops norvegicus burrows in underwater television surveys through machine learning
Papini O., Cecapolli E., Domenichetti F., Martinelli M., Pieri G., Reggiannini M., Zacchetti L.
The paper introduces computer vision methods for automating the detection, recognition, and classification of Nephrops norvegicus burrows in underwater videos. This approach aims to improve accuracy, reduce human errors, and standardize the current manual video analysis process. By using machine learning techniques, the system can automatically process video streams and detect N. norvegicus burrow openings on the seabed. The work also explores the use of data augmentation algorithms to extend the annotated data set, enhancing the performance of the automated system compared to the original manual annotations.Source: PATTERN RECOGNITION AND IMAGE ANALYSIS, vol. 34 (issue 4), pp. 1030-1036
DOI: 10.1134/s1054661824701062
Project(s): NAUTILOS
Metrics:
Papini O., Cecapolli E., Domenichetti F., Martinelli M., Pieri G., Reggiannini M., Zacchetti L.
The paper introduces computer vision methods for automating the detection, recognition, and classification of Nephrops norvegicus burrows in underwater videos. This approach aims to improve accuracy, reduce human errors, and standardize the current manual video analysis process. By using machine learning techniques, the system can automatically process video streams and detect N. norvegicus burrow openings on the seabed. The work also explores the use of data augmentation algorithms to extend the annotated data set, enhancing the performance of the automated system compared to the original manual annotations.Source: PATTERN RECOGNITION AND IMAGE ANALYSIS, vol. 34 (issue 4), pp. 1030-1036
DOI: 10.1134/s1054661824701062
Project(s): NAUTILOS
Metrics:
See at:
CNR IRIS | link.springer.com | CNR IRIS | CNR IRIS
2025
Other
Open Access
A survey on SAR ship classification using deep learning
Awais Ch Muhammad, Reggiannini M., Moroni D., Salerno E.
Deep learning (DL) has emerged as a powerful tool for Synthetic Aperture Radar (SAR) ship classification. This survey comprehensively analyzes the diverse DL techniques employed in this domain. We identify critical trends and challenges, highlighting the importance of integrating handcrafted features, utilizing public datasets, data augmentation, fine-tuning, explainability techniques, and fostering interdisciplinary collaborations to improve DL model performance. This survey establishes a first-of-its-kind taxonomy for categorizing relevant research based on DL models, handcrafted feature use, SAR attribute utilization, and the impact of fine-tuning. We discuss the methodologies used in SAR ship classification tasks and the impact of different techniques. Finally, the survey explores potential avenues for future research, including addressing data scarcity, exploring novel DL architectures, incorporating interpretability techniques, and establishing standardized performance metrics. By addressing these challenges and leveraging advancements in DL, researchers can contribute to developing more accurate and efficient ship classification systems, ultimately enhancing maritime surveillance and related applications.DOI: 10.48550/arxiv.2503.11906
DOI: https://doi.org/10.48550/arxiv.2503.11906
Metrics:
Awais Ch Muhammad, Reggiannini M., Moroni D., Salerno E.
Deep learning (DL) has emerged as a powerful tool for Synthetic Aperture Radar (SAR) ship classification. This survey comprehensively analyzes the diverse DL techniques employed in this domain. We identify critical trends and challenges, highlighting the importance of integrating handcrafted features, utilizing public datasets, data augmentation, fine-tuning, explainability techniques, and fostering interdisciplinary collaborations to improve DL model performance. This survey establishes a first-of-its-kind taxonomy for categorizing relevant research based on DL models, handcrafted feature use, SAR attribute utilization, and the impact of fine-tuning. We discuss the methodologies used in SAR ship classification tasks and the impact of different techniques. Finally, the survey explores potential avenues for future research, including addressing data scarcity, exploring novel DL architectures, incorporating interpretability techniques, and establishing standardized performance metrics. By addressing these challenges and leveraging advancements in DL, researchers can contribute to developing more accurate and efficient ship classification systems, ultimately enhancing maritime surveillance and related applications.DOI: 10.48550/arxiv.2503.11906
DOI: https://doi.org/10.48550/arxiv.2503.11906
Metrics:
See at:
arXiv.org e-Print Archive | CNR IRIS | doi.org | CNR IRIS
2025
Other
Open Access
Guidelines for the annotation of Nephrops norvegicus UWTV videos
Papini O., Cecapolli E., Domenichetti F., Martinelli M., Pieri G., Reggiannini M., Zacchetti L.
This document describes a methodology conceived to create ground truth datasets that may be exploited in the implementation of object detection and classification algorithms tailored on the Nephrops norvegicus. In fact, supervised machine learning algorithms usually require considerable amounts of annotated data to carry out the training stage. The greater the size of the annotated dataset, the stronger the required effort from the annotators.DOI: 10.32079/isti-tr-2025/009
DOI: 10.5281/zenodo.14973160
DOI: 10.5281/zenodo.14973159
Project(s): NAUTILOS
Metrics:
Papini O., Cecapolli E., Domenichetti F., Martinelli M., Pieri G., Reggiannini M., Zacchetti L.
This document describes a methodology conceived to create ground truth datasets that may be exploited in the implementation of object detection and classification algorithms tailored on the Nephrops norvegicus. In fact, supervised machine learning algorithms usually require considerable amounts of annotated data to carry out the training stage. The greater the size of the annotated dataset, the stronger the required effort from the annotators.DOI: 10.32079/isti-tr-2025/009
DOI: 10.5281/zenodo.14973160
DOI: 10.5281/zenodo.14973159
Project(s): NAUTILOS
Metrics:
2025
Journal article
Open Access
SAR-to-Infrared domain adaptation for maritime surveillance with limited data
Awais Ch Muhammad, Reggiannini M., Moroni D., Galdelli A.
Deep Learning (DL) algorithms need extensive amounts of data for classification tasks, which can be costly in specialized fields like maritime monitoring. To address data scarcity, we propose a fine-tuning approach leveraging complementary Infrared (IR) and Synthetic Aperture Radar (SAR) datasets. We evaluated our method using the ISDD, HRSID, and FuSAR datasets, employing VGG16 as a shared backbone integrated with Faster R-CNN (for ship detection) and a three-layer classifier (for ship classification). The results showed significant improvements in IR ship detection (mAP: +20%; Recall: +17%) and modest but consistent gains in SAR ship detection tasks (F1-score: +3%, Recall: +1%, mAP: +1%). Our findings highlight the effectiveness of domain adaptation in improving DL’s performance under limited data conditions.Source: PROCEEDINGS, vol. 129 (issue 1)
DOI: 10.3390/proceedings2025129066
Metrics:
Awais Ch Muhammad, Reggiannini M., Moroni D., Galdelli A.
Deep Learning (DL) algorithms need extensive amounts of data for classification tasks, which can be costly in specialized fields like maritime monitoring. To address data scarcity, we propose a fine-tuning approach leveraging complementary Infrared (IR) and Synthetic Aperture Radar (SAR) datasets. We evaluated our method using the ISDD, HRSID, and FuSAR datasets, employing VGG16 as a shared backbone integrated with Faster R-CNN (for ship detection) and a three-layer classifier (for ship classification). The results showed significant improvements in IR ship detection (mAP: +20%; Recall: +17%) and modest but consistent gains in SAR ship detection tasks (F1-score: +3%, Recall: +1%, mAP: +1%). Our findings highlight the effectiveness of domain adaptation in improving DL’s performance under limited data conditions.Source: PROCEEDINGS, vol. 129 (issue 1)
DOI: 10.3390/proceedings2025129066
Metrics:
See at:
CNR IRIS | www.mdpi.com | CNR IRIS
2024
Journal article
Open Access
Remote sensing for maritime traffic understanding
Reggiannini M, Salerno E, Bacciu C, D'Errico A, Lo Duca A, Marchetti A, Martinelli M, Mercurio C, Mistretta A, Righi M, Tampucci M, Di Paola C
The capability of prompt response in case of critical circumstances occurring within a maritime scenario depends on the awareness level of the competent authorities. From this perspective a quick and integrated surveillance service represents a tool of utmost importance. This is even more true when the main purpose is to tackle illegal activities such as smuggling, waste flooding or malicious vessel trafficking. This work presents an improved version of the OSIRIS system, a previously developed ICT framework devoted to understand the maritime vessel traffic through the exploitation of optical and radar data captured by satellite imaging sensors. A number of dedicated processing units are cascaded with the objective of i) detecting the presence of vessel targets in the input imagery, ii) estimating the vessel types on the basis of their geometric and scatterometric features, iii) estimating the vessel kinematics, iv) classifying the navigation behaviour of the vessel and predicting its route and, eventually, v) integrating the several outcomes within a webGIS interface to easily assess the traffic status inside the considered area. The entire processing pipeline has been tested on satellite imagery captured within the Mediterranean Sea or extracted from public, annotated data sets.Source: REMOTE SENSING (BASEL), vol. 16 (issue 3)
DOI: 10.3390/rs16030557
Metrics:
Reggiannini M, Salerno E, Bacciu C, D'Errico A, Lo Duca A, Marchetti A, Martinelli M, Mercurio C, Mistretta A, Righi M, Tampucci M, Di Paola C
The capability of prompt response in case of critical circumstances occurring within a maritime scenario depends on the awareness level of the competent authorities. From this perspective a quick and integrated surveillance service represents a tool of utmost importance. This is even more true when the main purpose is to tackle illegal activities such as smuggling, waste flooding or malicious vessel trafficking. This work presents an improved version of the OSIRIS system, a previously developed ICT framework devoted to understand the maritime vessel traffic through the exploitation of optical and radar data captured by satellite imaging sensors. A number of dedicated processing units are cascaded with the objective of i) detecting the presence of vessel targets in the input imagery, ii) estimating the vessel types on the basis of their geometric and scatterometric features, iii) estimating the vessel kinematics, iv) classifying the navigation behaviour of the vessel and predicting its route and, eventually, v) integrating the several outcomes within a webGIS interface to easily assess the traffic status inside the considered area. The entire processing pipeline has been tested on satellite imagery captured within the Mediterranean Sea or extracted from public, annotated data sets.Source: REMOTE SENSING (BASEL), vol. 16 (issue 3)
DOI: 10.3390/rs16030557
Metrics:
See at:
Remote Sensing | Remote Sensing | IRIS Cnr | IRIS Cnr | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
A mesoscale events classifier for sea surface temperature data
Reggiannini M., Papini O., Pieri G.
The identification of mesoscale phenomena, such as upwelling, countercurrents and filaments, is an important task for oceanographers. Indeed, the occurrence of such processes involves variations in the density of nutrients which, in turn, influences the biological parameters of the habitat. In this work, we describe a novel method for an automatic classification system, the Mesoscale Events Classifier (MEC), dedicated to recognising marine mesoscale events. MEC is devoted to the study of these phenomena through the analysis of Sea Surface Temperature (SST) images captured by satellite missions.Source: MISCELLANEA INGV, vol. 80, pp. 317-319. Bergen, Norvegia, 27-29/05/2024
DOI: 10.13127/misc/80/122
DOI: https://commons.datacite.org/doi.org/10.13127/misc/80/122
Project(s): NAUTILOS
Metrics:
Reggiannini M., Papini O., Pieri G.
The identification of mesoscale phenomena, such as upwelling, countercurrents and filaments, is an important task for oceanographers. Indeed, the occurrence of such processes involves variations in the density of nutrients which, in turn, influences the biological parameters of the habitat. In this work, we describe a novel method for an automatic classification system, the Mesoscale Events Classifier (MEC), dedicated to recognising marine mesoscale events. MEC is devoted to the study of these phenomena through the analysis of Sea Surface Temperature (SST) images captured by satellite missions.Source: MISCELLANEA INGV, vol. 80, pp. 317-319. Bergen, Norvegia, 27-29/05/2024
DOI: 10.13127/misc/80/122
DOI: https://commons.datacite.org/doi.org/10.13127/misc/80/122
Project(s): NAUTILOS
Metrics:
2024
Conference article
Open Access
Advancing sustainability: research initiatives at the Signals and Images Lab
Bruno A., Caudai C., Conti F., Leone G. R., Magrini M., Martinelli M., Moroni D., Muhammad A. Ch, Papini O., Pascali M. A., Pieri G., Reggiannini M., Righi M., Salerno E., Scozzari A., Tampucci M.
In this paper, we aim to briefly survey the relations of the work conducted at the Signals and Images Lab of CNR-ISTI, Pisa, with the themes of sustainability. We explore both the broader implications and the application-specific aspects of our work, highlighting references to published research and collaborative projects undertaken with key stakeholders and industrial partners.Source: CEUR WORKSHOP PROCEEDINGS, vol. 3762, pp. 499-504. Napoli, Italy, 29-30/05/2024
Bruno A., Caudai C., Conti F., Leone G. R., Magrini M., Martinelli M., Moroni D., Muhammad A. Ch, Papini O., Pascali M. A., Pieri G., Reggiannini M., Righi M., Salerno E., Scozzari A., Tampucci M.
In this paper, we aim to briefly survey the relations of the work conducted at the Signals and Images Lab of CNR-ISTI, Pisa, with the themes of sustainability. We explore both the broader implications and the application-specific aspects of our work, highlighting references to published research and collaborative projects undertaken with key stakeholders and industrial partners.Source: CEUR WORKSHOP PROCEEDINGS, vol. 3762, pp. 499-504. Napoli, Italy, 29-30/05/2024
See at:
ceur-ws.org | CNR IRIS | CNR IRIS
2024
Conference article
Restricted
Deep learning for SAR ship classification: focus on unbalanced datasets and inter-dataset generalization
Awais C. M., Reggiannini M.
Detection and recognition of vessel targets at sea are tasks of paramount relevance for maritime monitoring purposes. A possible approach to pursue these objectives consists in acquiring and processing Synthetic Aperture Radar (SAR) data related to a given area of interest. Classically, the detection part can be implemented by exploiting statistical properties of the signal to decide whether an image area belongs to background clutter or to a ship (e.g. Constant False Alarm Rate based algorithms). Successively, discriminant features referring to the detected object can be extracted and later fed to a classifier to decide the membership category of the considered target. Recently, thanks to the development of algorithms based on deep neural network architectures, object detection and recognition experienced an unprecedented boost in the observed performances. This work, mainly motivated by the exploration of these novel approaches to the identification of vessel targets, focuses on the analysis of five different deep learning architectures (CNN, pre-trained and non-pretrained versions of ResNet50 and VGG16) trained on two public SAR vessel datasets (OpenSARShip and Fusar). To address the data quantity limitation, a third dataset was created by merging both datasets.DOI: 10.1109/iceaa61917.2024.10701968
Project(s): National Biodiversity Future Center
Metrics:
Awais C. M., Reggiannini M.
Detection and recognition of vessel targets at sea are tasks of paramount relevance for maritime monitoring purposes. A possible approach to pursue these objectives consists in acquiring and processing Synthetic Aperture Radar (SAR) data related to a given area of interest. Classically, the detection part can be implemented by exploiting statistical properties of the signal to decide whether an image area belongs to background clutter or to a ship (e.g. Constant False Alarm Rate based algorithms). Successively, discriminant features referring to the detected object can be extracted and later fed to a classifier to decide the membership category of the considered target. Recently, thanks to the development of algorithms based on deep neural network architectures, object detection and recognition experienced an unprecedented boost in the observed performances. This work, mainly motivated by the exploration of these novel approaches to the identification of vessel targets, focuses on the analysis of five different deep learning architectures (CNN, pre-trained and non-pretrained versions of ResNet50 and VGG16) trained on two public SAR vessel datasets (OpenSARShip and Fusar). To address the data quantity limitation, a third dataset was created by merging both datasets.DOI: 10.1109/iceaa61917.2024.10701968
Project(s): National Biodiversity Future Center
Metrics:
2024
Other
Open Access
Imbalanced datasets through the lens of transfer-learning
Awais Ch Muhammad., Reggiannini M.
Data scarcity and class imbalance hinder deep learning for tasks like SAR ship classification. This work investigates how TRANSFER-LEARNING and DATA MERGING techniques can significantly improve the performance of deep learning models for class imbalanced datasets.Project(s): National Biodiversity Future Center
Awais Ch Muhammad., Reggiannini M.
Data scarcity and class imbalance hinder deep learning for tasks like SAR ship classification. This work investigates how TRANSFER-LEARNING and DATA MERGING techniques can significantly improve the performance of deep learning models for class imbalanced datasets.Project(s): National Biodiversity Future Center
See at:
CNR IRIS | www.nbfc.it | CNR IRIS
2024
Conference article
Open Access
Evaluating the Impact of fine-tuning on deep learning models for SAR ship classification
Awais Ch Muhammad, Reggiannini M.
Tasks like SAR ship classification suffer in deep learning due to data scarcity and class imbalance. To overcome these challenges, techniques like fine-tuning and data merging can play a vital role in the performance of a deep learning model. This study evaluates the effect of fine tuning on 5 different deep learning models.Project(s): National Biodiversity Future Center
Awais Ch Muhammad, Reggiannini M.
Tasks like SAR ship classification suffer in deep learning due to data scarcity and class imbalance. To overcome these challenges, techniques like fine-tuning and data merging can play a vital role in the performance of a deep learning model. This study evaluates the effect of fine tuning on 5 different deep learning models.Project(s): National Biodiversity Future Center
See at:
CNR IRIS | www.nbfc.it | CNR IRIS