Sludge Snap: A machine learning approach to faecal sludge characterisation in the field скачать в хорошем качестве

Sludge Snap: A machine learning approach to faecal sludge characterisation in the field

This video describes the current status of our Sludge Snap app. Based on smart phone pictures of faecal sludge, the Sludge Snap app uses a machine learning approach to estimate characteristics of faecal sludge in the field, and to help fill a gap when there is limited access to laboratories. The current study was a proof of concept based on samples from 420 different onsite containments in Lusaka, Zambia. We found that TS, NH4-N, and dewatering performance could be predicted with machine learning models based on the smart phone pictures, and also based on probe readings (pH and conductivity). The video also explains the necessary next steps in order to field test and validate the results, with the goal of being able to make real-time predictions in global applications. This presentation was part of the 42nd WEDC International Conference.