This repo gathers open-source sonar image resources for classification, detection, and segmentation.
📣More open resources links are welcome😃. Let's build a better open-source environment for sonar image recognition community together!❤️Please feel free to contribute!
1.NKSID
Description: A newly created forward-looking sonar image recognition benchmark, named NanKai Sonar Image Dataset (NKSID). This dataset contains 2617 images from 8 categories, with labels showing a natural long tail distribution.
Description: This open dataset contains 385 ship images and 62 airplane images, and can be used for acedemic purpose.
3.marine-debris-fls-datasets-turntable-cropped
Description: Marine-Debris Datasets captured with a Forward-Looking Sonar in a water tank and turntable using an ARIS Explorer 3000 Sonar. The turntable dataset is captured with the sonar in a fixed position and pose, while objects are placed in a rotating turntable, allowing to capture a full yaw rotation of each object. For now only a classification task is available.
1.Forward-looking-Sonar-Detection-Dataset
Description: This forward-looking sonar(FLS) detection dataset including underwater victim, boat, and plane data. All the FLS data is obtained by Oculus M1200d sonar for academic research.
2.SCTD: Sonar Common Target Detection Dataset
Description: It collected 497 high resolution images obtained by side-scan sonars, forward-looking sonars, and (interferometric) synthetic aperture sonars, on which 596 targets are included. (The average number of targets per image is 1.2.)
3.UATD
Description: The Pengcheng Laboratory provided this underwater sonar image dataset, comprising 7600 sonar images, and made it available as scientific data. This dataset includes three categories of regular geometric structure objects (cube, ball, and cylinder) and seven other types of underwater objects (human body, metal bucket, tyre, square cage, circle cage, plane, and ROV).
4.LSOD
Description: This is a repo for small target sonar image detection. Full dataset will be available after their paper being accepted.
1.Forward-looking-Sonar-Segmentation-Dataset
Description: This forward-looking sonar(FLS) segmentation dataset including underwater boat and plane data. All the FLS data is obtained by Oculus M1200d sonar for academic research.
2.marine-debris-fls-datasets-watertank-segmentation
Description: The dataset consists of images obtained from a forward looking SONAR and the corresponding semantic labels for 11 classes (except background).
Description: This is the official repository for the Self-supervised Learning for Sonar Image Classification paper published at CVPR2022 workshop.
Description: This repo shows the source code of IEEE TGRS 2022 article: Sonar Images Classification While Facing Long-Tail and Few-Shot.
Description: This repo contains the implementation code for the ESWA 2024 article Open-set recognition with long-tail sonar images.
This is a repo for Kesci underwater object detection algorithm contest. (Readme file is in Chinese)
This is the official codes for IEEE TGRS 2021 paper Self-trained detection detectors are implemented on small-scale Radar and Sonar detection datasets.