Deep Learning on SAR Imagery: Transfer Learning Versus Randomly Initialized Weights

Morteza Karimzadeh, Rafael Pires de Lima
10/26/2023
cs.CVeess.IV

No official code found. Best implementation: geohai/sea-ice-segment (18 stars, last commit 2025-01-06).

Abstract

Deploying deep learning on Synthetic Aperture Radar (SAR) data is becoming more common for mapping purposes. One such case is sea ice, which is highly dynamic and rapidly changes as a result of the combined effect of wind, temperature, and ocean currents. Therefore, frequent mapping of sea ice is necessary to ensure safe marine navigation. However, there is a general shortage of expert-labeled data to train deep learning algorithms. Fine-tuning a pre-trained model on SAR imagery is a potential solution. In this paper, we compare the performance of deep learning models trained from scratch using randomly initialized weights against pre-trained models that we fine-tune for this purpose. Our results show that pre-trained models lead to better results, especially on test samples from the melt season.

View on arXivView PDF

Code Implementations(1)

geohai/sea-ice-segmentCommunity100%

Sea ice segmentation using synthetic aperture radar and convolutional neural networks

182PythonDec 6, 20221 years ago

Similar Papers

Analytical Logit Scaling for High-Resolution Sea Ice Topology Retrieval from Weakly Labeled SAR Imagery

86% match2026

Partial Label Learning with Focal Loss for Sea Ice Classification Based on Ice Charts

86% match1 repo2024

IceDiff: High Resolution and High-Quality Arctic Sea Ice Forecasting with Generative Diffusion Prior

86% match1 repo102026

Comparison of Cross-Entropy, Dice, and Focal Loss for Sea Ice Type Segmentation

85% match1 repo182023

Surrogate Modelling for Sea Ice Concentration using Lightweight Neural Ensemble

85% match2023

MT-IceNet -- A Spatial and Multi-Temporal Deep Learning Model for Arctic Sea Ice Forecasting

85% match1 repo202023

Cite this paper

@article{karimzadeh2023deep,
  title  = {Deep Learning on SAR Imagery: Transfer Learning Versus Randomly Initialized Weights},
  author = {Morteza Karimzadeh and Rafael Pires de Lima},
  year   = {2023},
  eprint = {2310.17126},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2310.17126v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2310.17126.svg)](https://codeofpaper.com/paper/2310.17126)

Discussion