Enhancing the Quality of Available Maps in Palestine using Deep Learning Model

Local Chapter Palestine Chapter

Coordinated by ,

Status: Completed

Project Duration: 15 Jun 2023 - 10 Aug 2023

Open Source resources available from this project

Project background.

The lack of precise maps for Palestinian territories is a complex issue that is rooted in political, economic, and historical factors. The BBC article by Christopher Giles & Jack Goodman, claims that:

“Much of both Israel and the Palestinian territories appear on Google Earth as low-resolution satellite imagery, even though higher-quality images are available from satellite companies.”

Blurry satellite images affected the correctness of roads definitions so Google maps users in Palestine usually are provided wrong directions to places, and even may end their journey in the middle of desert, segregation wall and military bases or other dangerous places.

Since 1997, the US government had restricted American companies from providing high-quality satellite images of the Palestinian territories due to Israeli security concerns, resulting in low-quality images on mapping applications. The Kyl-Bingaman Amendment (KBA) had limited image quality to the extent that objects smaller than a car would be challenging to identify. The amendment was dropped in July 2020 after increasing pressure from academics and the availability of higher-resolution images from non-US providers.

“As a result of recent changes to US regulations, the imagery of Israel and Gaza is being provided at 40cm resolution,” Maxar, one of the biggest providers of satellite images said in a statement. However, Google has not shared any plans to improve its satellite imagery.

As local communities rely heavily on free-to-use mapping software, they still lack access to high-resolution up-to-date maps.

Project plan.

  • Week 1

    Research previous work and Data Collection

  • Week 2

    Data Collection

  • Week 3

    EDA

  • Week 4

    Data Preprocessing

  • Week 5

    Model Development and Training

  • Week 6

    Model Finetuning and interpretation

  • Week 7

    Deployment

  • Week 8

    Report

Learning outcomes.

– GANs
– Computer vision
– Deep learning
– Working with satellite images
– Working through the entire ML pipeline

Share project on: