FICS 2024 Registrations are now open! Register Now
Idea - FICS

SEMANTIC SEGMENTATION FOR GLACIERS USING SATELLITE IMAGERY

Team and Contact Details

Student Name School Degree Year Email
Hamza AmjadSEECSUndergraduateFourth[email protected]
Ahmad Hassan IsmailSEECSUndergraduateFourth[email protected]

Inter School Idea ? No
Do you need expertises from another area: No
If Yes please provide details of expertises you need:

Idea Details

Idea Name: SEMANTIC SEGMENTATION FOR GLACIERS USING SATELLITE IMAGERY
Slogan: Be vanish
Supervisor Name: Dr Muhammad Imran Malik
Supervisor Designation: Professor
Supervisor School: SEECS
Supervisor Department: Deep Learning - AI
Contact number: +923335225189
Email ID: [email protected]
Abstract:
    Our main focus will be on the design and implementation of a model to extract semantic maps of waters, ice, glaciers, land, and clouds to track the water scarcity in Pakistan. We aim for designing a deep learning-based algorithm for semantic segmentation of glaciers from satellite imagery. We perform various analytics tasks on the segmented data.
What is the unmet need in society that your idea will fulfill ?
    • There is a lot of publicly available dataset. But, there is no satellite imagery labeled-dataset for Pakistan. • There is no deep learning model that helps Pakistan in extracting the information from satellite images. • There is no publicly available segmented data for the glacier and hilly areas
Who needs it ? How many would benefit ?
   • National Disaster Management Authority • Water Supply and Sewerage Authority • Government of Pakistan The Government of Pakistan is continuously trying to overcome this problem, especially in winters. So, it makes it difficult to go there in person to resolve the issues.
How will the solution works
    In addition to U-Net Architecture, a new architecture deeplabv3 is used that has more performance, computation, and accuracy as compared to other ones. The encoder consisting of pre-trained CNN models is used to get encoded feature maps of the input image, and the decoder reconstructs output, from the essential information extracted by the encoder, using up-sampling. It includes Atrous Separable Convolution for each input channel along with point-wise convolution.
Who are your competitors ? How is your solution different
    From the start we were sure to come up with an idea that would matter for the people of Pakistan. A method that provides constant feedback about the water level of Pakistan lakes, valleys, mountainous areas, the Arabian Sea, and some other watered parts of Pakistan.
Status: new
Entry Date & Time: 2021-12-23 (1901)