Today all students in the Hack4Good 8-week pro-bono machine learning program presented their final work.
BlueGreen Labs supervised the program on behalf of the IFPRI Picture Based Insurance (PBI) program and ACRE Africa. The students who worked on the the IFPRI challenge created a multi-modal machine learning pipeline to classify crop damages from the combination of both weather data and cellphone (PBI) images.
The students presented their work showing that damage could be accurately classified in most cases. Contrary to expectations additional weather data did not provided more accurate results. The students hypothesized that a more complex LSTM based approach (a different model architecture) could increase accuracy. However, time constraints limited their ability to explore more options.
Once more, congratulations to all the students for their contributions ensuring food security solutions using machine learning.
