Best practices in convolutional networks for forward-looking sonar image recognition
Open accessAbstract
Convolutional Neural Networks (CNN) have revolutionized perception for color images, and their application to sonar images has also obtained good results. But in general CNNs are difficult to train without a large dataset, need manual tuning of a considerable number of hyperparameters, and require many careful decisions by a designer. In this work, we evaluate three common decisions that need to be made by a CNN designer, namely the performance of transfer learning, the effect of object/image size and the relation between training set size. We evaluate three CNN models, namely one based on LeNet, and two based on the Fire module from SqueezeNet. Our findings are: Transfer learning with an SVM works very well, even when the train and transfer sets have no classes in common, and high classification performance can be obtained even when the target dataset is small. The ADAM optimizer combined with Batch Normalization can make a high accuracy CNN classifier, even with small image sizes (16 pixels). At least 50 samples per class are required to obtain 90% test accuracy, and using Dropout with a small dataset helps improve performance, but Batch Normalization is better when a large dataset is available.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
CNN models are largely invariant to object size.
“Additionally, CNN models are largely invariant to the object size”
From A Survey of Underwater Acoustic Data Classification Methods Using Deep Learning for Shoreline Surveillance (2022) · cites this paper as [131]
The Conclusions recommend a classic CNN model for high classification performance that is highly invariant to object size. But Section V.B reports size-dependent accuracy for TinyNet and FireNet, and for ClassicCNN when trained with SGD. The claim overgeneralizes a result that depends on the model and training setup; page numbers are not provided in our copy. (LLM, conf=0.87)
Experimental — accuracy not yet measuredA combination of the ADAM optimiser with Dropout regularisation is a good strategy to cope with the small sample size problem.
“and finally, a combination of ADAM optimiser with Dropout regularisation was shown to be a good strategy to cope with the small sample size problem.”
From A Survey of Underwater Acoustic Data Classification Methods Using Deep Learning for Shoreline Surveillance (2022) · cites this paper as [131]
CNNs, even when trained with distinct datasets from the target, are capable of learning relevant features, improving classification accuracy using transfer learning.
“Results show that CNNs (even when trained with distinct datasets from the target) are capable of learning relevant features, improving the accuracy of classification using transfer learning.”
From A Survey of Underwater Acoustic Data Classification Methods Using Deep Learning for Shoreline Surveillance (2022) · cites this paper as [131]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · see citing works
CNN Features off-the-shelf: an Astounding Baseline for Recognition
Razavian, A. S., Azizpour, H., Sullivan, J., & Carlsson, S. (2014) · arXiv · doi:10.48550/arxiv.1403.6382 · Cited by 607 · see citing works
Flagged first, then by citations
Research works printed but not linked: 11
- J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, "Imagenet: A large-scale hierarchical image database," in Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on. IEEE, 2009, pp. 248-255.
- F. N. Iandola, M. W. Moskewicz, K. Ashraf, S. Han, W. J. Dally, and K. Keutzer, "Squeezenet: Alexnet-level accuracy with 50x fewer parameters and < 1mb model size," arXiv preprint arXiv:1602.07360, 2016. · arXiv
- M. Valdenegro-Toro, "Real-time convolutional networks for sonar image classification in low-power embedded systems," in To appear in Proceedings of the European Symposium in Artificial Neural Networks, 2017.
- J. Yosinski, J. Clune, Y. Bengio, and H. Lipson, "How transferable are features in deep neural networks?" in Advances in neural information processing systems, 2014, pp. 3320-3328.
- Y. Pailhas, Y. Petillot, and C. Capus, "High-resolution sonars: what resolution do we need for target recognition?" EURASIP Journal on Advances in Signal Processing, vol. 2010, no. 1, p. 205095, 2010.
- D. Mishkin, N. Sergievskiy, and J. Matas, "Systematic evaluation of cnn advances on the imagenet," arXiv preprint arXiv:1606.02228, 2016. · arXiv
- I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. MIT Press, 2016, http://www.deeplearningbook.org.
- Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998.
- A. Razavian, H. Azizpour, J. Sullivan, and S. Carlsson, "Cnn features off-the-shelf: an astounding baseline for recognition," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2014, pp. 806-813.
- S. Ioffe and C. Szegedy, "Batch normalization: Accelerating deep network training by reducing internal covariate shift," arXiv preprint arXiv:1502.03167, 2015. · arXiv
- N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhut- dinov, "Dropout: A simple way to prevent neural networks from overfitting," The Journal of Machine Learning Research, vol. 15, no. 1, pp. 1929-1958, 2014.
Cited by
Sorted by citations
Domingos, L. C. F., Santos, P. E., Skelton, P. S. M., Brinkworth, R. S. A., & Sammut, K. (2022) · Sensors · doi:10.3390/s22062181 · Cited by 93 · see citing works
“The MDT dataset was used in [131] to evaluate the performance of transfer learning using the recognition accuracy as a function of the feature vector size, the object size as well as the training set size.”
BackgroundMethodPre-trained Models for Sonar Images
Valdenegro-Toro, M., Preciado-Grijalva, A., & Wehbe, B. (2021) · OCEANS 2021: San Diego – Porto · doi:10.23919/oceans44145.2021.9705825 · Cited by 21 · see citing works
Self-citation
- Authors
- Valdenegro-Toro, M.
- Venue
- OCEANS 2017 - Aberdeen
- Publisher
- IEEE
- Published
- 2017-06
- Pages
- 1-9
- OpenAlex
- W2751875176
- Cited by
- 30
- References
- 2
- Citations by year
- 2025: 2 · 2024: 5 · 2023: 5 · 2022: 6 · 2021: 4 · 2020: 4 · 2019: 1 · 2018: 3
- FWCI
- 0.64
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- MTS/IEEE OCEANS Conference (OCEANS) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Earth and Planetary Sciences › Oceanography › Underwater Acoustics Research
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- convolutional neural networks, forward-looking sonar, transfer learning, SqueezeNet, LeNet, fire module, limited training data, image size, batch normalization, Adam optimizer, dropout, support vector machine, underwater target classification
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
