michaelhasey/Maasai-Skywatch

Using Yolo-V3 to detect burned Maasai villages & provide population counts. # Maasai Skywatch ### Detecting Illegal Village Burns and Other Spatial Injustices Michael Hasey, Luke McKinley, Rhys Broussard ## Summary Maasai Skywatch is a proposed online tool that uses publicly available satellite imagery and sophisticated object-detecting algorithms to identify and locate illegal village burns and land-related injustices against the Maasai people. ## Colab Notebook A colab notebook has been provided to recreate the following experiment ``` maasai-village-notebook.ipynb ``` ## Project Page An expanded description of Maasai Skywatch can be found on its project page here: ``` michaelhasey.com ``` ## Table of Contents - Background - Dataset - Model - Performance - References - Citation ## Background A semi-nomadic ethnic group, the Maasai are based in the Great Rift Valley of East Africa, between Northern Tanzania and Southern and Central Kenya. With a population of almost 2 million, they have inhabited the region for centuries, moving their herds of cattle with the seasons. Government officials and foreign companies in Tanzania and Kenya are using ecotourism and conservation laws to displace indigenous Maasai people, evicting them and denying them access to watering holes and vital grazing for their livestock. We use computer vision to provide the Maasai with a tool to quickly identify illegal village burns, forced migrations, & potential land grabs. The advantage of AI assistance is its speed and accuracy. While it may take multiple weeks for NGO’s, word of mouth and manual village visits to determine the extent of damage, a combination of AI and “real time” satellite imagery may help to reduce this time to mere hours or minutes. By providing quick and accurate counts and metrics of land-based injustices with soon after they occur, Maasai Skywatch’s AI technology can one day provide the robust, unbiased, and hard evidence needed to fight many of these social injustices and protect Maasai l …