https://doi.org/10.1140/epjp/s13360-025-06498-4
Regular Article
HydR-CNN: advancing underwater object detection using a multi-stage framework with hybrid R-CNN and pyramid vision transformer with augmented convolution
School of Computer Science Engineering and Technology, Bennett University, Greater Noida, Uttar Pradesh, India
Received:
8
November
2024
Accepted:
30
May
2025
Published online:
19
June
2025
Underwater object detection is an inevitable technology in marine science, environmental survey, and underwater robotics where reliable identification and classification of the objects are mandatory for research, environmental management, and search and rescue operations. This field has many difficulties due to small light conditions, backgrounds, having various scales of objects, and also due to the required effective methods of features extraction. In response to these difficulties, new method called HydR-CNN has been established as it uses advanced strategies to improve the detection. In the HydR-CNN method, multi-stage feature extraction is used with the help of the Pyramid Vision Transformer that enhances the identification of clear and accurate features for low visibility conditions. Also, the Hierarchical Dynamic R-CNN component controls receptive fields and hierarchical layers for manner with objects in diverse scales and intricacies. New strategies termed multi-scale classification and bounding box refinement are proposed to enhance the object localization and to minimize the false detections. A test on Deep Fish, RUOD, and DUO, three new underwater object detection datasets reveal that HydR-CNN offers a better TPR, FPR, precision, recall and F1-score as compared to earlier methods. Further analysis of the visualization outcome again demonstrates that our model achieves promising confidence scores across multiple classes, which again corroborates the model’s reliability in detecting and identifying various underwater objects. Thus, proving HydR-CNN to be a robust and effective solution for the complex detection of underwater objects.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Società Italiana di Fisica and Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
