Globalization has led to a radical transformation of the trade in food and other agricultural products over the last decades, with repercussions on economies around the world. In Sub-Saharan Africa, the export value of fruit, vegetables and flowers (high-value agricultural products) has increased from 1.3 Billion US$ in 1998 to 8.8 Billion US$ in 2018. The change in the export structure has also brought about significant transformations is the organization of agricultural production, such as the rapid expansion of vertically-integrated horticultural production facilities (such as industrial-scale greenhouses).
While the emergence of modern agricultural supply chains is generally perceived as an opportunity for developing countries, the implications of horticultural exports for poverty and household welfare are still poorly understood. In order to be able to link poverty and welfare outcomes to the emergence of the modernization of agricultural supply chains, a better understanding is needed on the location and scale of these facilities, as well as to be able to understand when they were constructed.
The proposed student project involves geotagging existing greenhouses in Kenya (mostly in the cutflower industry), and to create routines of automatic detection of greenhouses in large areas using artificial intelligence/deep learning approaches.
This project will serve as a starting point for a cooperative project between GRS and AEP that will train similar models across Sub-Saharan Africa in order to estimate the welfare effects of the modernization of agricultural supply chains.
- Explore satellite imagery of different resolutions (Landsat, Sentinel 2 and Planet) in order to manually identify greenhouses in Kenya
- Build a ground truth suitable for training ML/DL models
- Set-up, train and evaluate multiple ML models, including DL, to perform greenhouse detection at the country level.
- Machine Learning (FTE-35306) or Deep Learning (GRS-34806)
- Programming in Python, Geo-Scripting
Theme(s): Sensing & measuring