Martin Simon
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Martin Simon has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Khaled Alhazmi
King Abdulaziz City Sci & Technol KACST, Commun & Informat Technol Res Inst, Natl Ctr Robot & Internet Things Technol, Riyadh 11442, Saudi Arabia
1 shared publication - Sander Rikka
Department of Marine Systems, Tallinn University of Technology, 12618 Tallinn, Estonia
1 shared publication - Saud Alrshoud
National Center for Robotics and Internet of Things Technology, Communication and Information Technologies Research Institute, King Abdulaziz City for Science and Technology—KACST, Riyadh 11442, Saudi Arabia
1 shared publication - Simon Wenkel
Marduk Technologies OÜ, 12618 Tallinn, Estonia
1 shared publication - Sven Nomm
Department of Software Science, Tallinn University of Technology, Tallinn, Estonia
1 shared publication - Tanel Liiv
Marduk Technologies OÜ, 12618 Tallinn, Estonia
1 shared publication
Latest Publications
- Open Access
- 9 November 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 16 (19391404) , 83-88
Application of the LSTM Models for Baltic Sea Wave Spectra Estimation
This article proposes to apply long-short-term memory (LSTM) deep learning models to transform Sentinel-1 A/B interferometric wide (IW) swath image data into the wave density spectrum. Although spectral wave estimation methods for synthetic aperture radar data have been developed, similar approaches for coastal areas have not received enough attention. Partially, this is caused by the lack of high-resolution wave-mode data, as well as the nature of wind waves that have more complicated backscattering mechanisms compared to the swell waves for which the aforementioned methods were developed. The application of the LSTM model has allowed the transformation of the Sentinel-1 A/B IW one-dimensional image spectrum into wave density spectra. The best results in the test dataset led to the mean Pearson's correlation coefficient 0.85 for the comparison of spectra and spectra. The result was achieved with the LSTM model using $VV$ and $VH$ polarization spectra fed into the model independently. Experiments with LSTM neural networks that classify images into wave spectra with the Baltic Sea dataset demonstrated promising results in cases where empirical methods were previously considered.
- 9 November 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 16 (19391404) , 83-88
Open AccessApplication of the LSTM Models for Baltic Sea Wave Spectra Estimation
This article proposes to apply long-short-term memory (LSTM) deep learning models to transform Sentinel-1 A/B interferometric wide (IW) swath image data into the wave density spectrum. Although spectral wave estimation methods for synthetic aperture radar data have been developed, similar approaches for coastal areas have not received enough attention. Partially, this is caused by the lack of high-resolution wave-mode data, as well as the nature of wind waves that have more complicated backscattering mechanisms compared to the swell waves for which the aforementioned methods were developed. The application of the LSTM model has allowed the transformation of the Sentinel-1 A/B IW one-dimensional image spectrum into wave density spectra. The best results in the test dataset led to the mean Pearson's correlation coefficient 0.85 for the comparison of spectra and spectra. The result was achieved with the LSTM model using $VV$ and $VH$ polarization spectra fed into the model independently. Experiments with LSTM neural networks that classify images into wave spectra with the Baltic Sea dataset demonstrated promising results in cases where empirical methods were previously considered.
- Open Access
Confidence Score: The Forgotten Dimension of Object Detection Performance Evaluation
When deploying a model for object detection, a confidence score threshold is chosen to filter out false positives and ensure that a predicted bounding box has a certain minimum score. To achieve state-of-the-art performance on benchmark datasets, most neural networks use a rather low threshold as a high number of false positives is not penalized by standard evaluation metrics. However, in scenarios of Artificial Intelligence (AI) applications that require high confidence scores (e.g., due to legal requirements or consequences of incorrect detections are severe) or a certain level of model robustness is required, it is unclear which base model to use since they were mainly optimized for benchmark scores. In this paper, we propose a method to find the optimum performance point of a model as a basis for fairer comparison and deeper insights into the trade-offs caused by selecting a confidence score threshold.
Open AccessConfidence Score: The Forgotten Dimension of Object Detection Performance Evaluation
When deploying a model for object detection, a confidence score threshold is chosen to filter out false positives and ensure that a predicted bounding box has a certain minimum score. To achieve state-of-the-art performance on benchmark datasets, most neural networks use a rather low threshold as a high number of false positives is not penalized by standard evaluation metrics. However, in scenarios of Artificial Intelligence (AI) applications that require high confidence scores (e.g., due to legal requirements or consequences of incorrect detections are severe) or a certain level of model robustness is required, it is unclear which base model to use since they were mainly optimized for benchmark scores. In this paper, we propose a method to find the optimum performance point of a model as a basis for fairer comparison and deeper insights into the trade-offs caused by selecting a confidence score threshold.