This project is a part of the collaboration between Water Point Data Exchange (wpdx) and DataKind Singapore.
WPDX aims to automate the water pump labeling when volunteers upload water point data to their website. In line with this, we'll be helping them create models which classify the water pump images and the water tech descriptions accordingly.
The following are the associated tasks:
- From the WPDX dataset, extract relevant water pump images (photo_lnk) to a file
- Associate the images (photo_lnk) with their labels (water_tech)
- Use the extracted data to create a model
- Create a sample app to demo the model to WPDX team.
Sample Image: https://tinyurl.com/water-pump-image
Sample Label: Hand Pump - Vergnet
Due to size constraints, we have uploaded only 100 images per class split into train and test data pre-downloaded for testing purposes. A bigger image dataset containing 1000 images per class (where available) split into train and test folders can be found in the link here. Do perform your model training on the train data set and evaluate it with the test data set.
Check contributing.md on how to contribute to the project.
