Core demands
At the same time of steadily promoting the poverty alleviation "decapitation" in Hanzhong, Shaanxi Province, based on the actual needs of work, it is seeking the optimal solution for accurate and rapid analysis of a large number of front-line pictures uploaded by "helping cadres".
Solution
The two deputy directors of the Information Management Software Center of Hanzhong Poverty Alleviation Office spent about three days to complete the annotation of 6000 pictures, and used EasyDL to train the housing quality identification model (to check whether it is adobe house, etc.) and the drinking water type identification model (to check whether it is difficult to draw well water, etc.). The accuracy rate reached more than 85% and the model is still being iterated. Then, the trained model was integrated into the poverty alleviation information management software of Hanzhong Poverty Alleviation Office to achieve image acquisition and AI auxiliary audit functions, and the recognition results of EasyDL were added to the poverty alleviation rating system.
Actual application effect of poverty alleviation information system: