CCESAR: SAR Images

Citation Author(s):
Vidhu
Arora
Graphics-Visualization-Computing Lab, International Institute of Information Technology Bangalore, Karnataka 560100, India
Shreyan
Gupta
Graphics-Visualization-Computing Lab, International Institute of Information Technology Bangalore, Karnataka 560100, India
Ananthakrishna
Kudupu
Graphics-Visualization-Computing Lab, International Institute of Information Technology Bangalore, Karnataka 560100, India
Aditya
Priyadarshi,
Graphics-Visualization-Computing Lab, International Institute of Information Technology Bangalore, Karnataka 560100, India
Aswathi
Mundayatt
Graphics-Visualization-Computing Lab, International Institute of Information Technology Bangalore, Karnataka 560100, India
Jaya
Sreevalsan-Nair
Graphics-Visualization-Computing Lab, International Institute of Information Technology Bangalore, Karnataka 560100, India
Submitted by:
Aswathi Mundaya...
Last updated:
Wed, 03/19/2025 - 11:13
DOI:
10.21227/q8aq-jy35
License:
Creative Commons Attribution
63 Views
Categories:
Image Processing
Computer Vision
Geoscience and Remote Sensing
Keywords:
Coastline extraction, classification, segmentation, Deep Learning, CNN model, U-Net model, SAR images, Compression format
0
0 ratings - Please login to submit your rating.
ACCESS DATASET CITE SHARE/EMBED

Abstract 

This dataset comprises 32-bit floating-point SAR images in TIFF format, capturing coastal regions. It includes corresponding ground truth masks that differentiate between land and water areas. The covered regions include the Netherlands, London, Ireland, Spain, France, Lisbon, the USA, India, Africa, and Italy. The SAR images were acquired in Interferometric Wide (IW) mode with dual polarization at a spatial resolution of 10m × 10m.

 

 

 

 

 

 

 

Instructions: 

The dataset is organized into two primary folders: train and test. Each of these folders contains two subfolders:  

natural: This subfolder contains data samples representing natural environments or features.

built: This subfolder contains data samples representing built or man-made environments or structures.  

The train folder is intended for model training, while the test folder is reserved for evaluating the model's performance on unseen data.

Dataset Files

LOGIN TO ACCESS DATASET FILES

 

QUESTIONS?