Suchergebnisse

Precise and comprehensive measurements of socioeconomic status are crucial for both academic investigations and policy-making. However, in developing countries such measures are available at the local household level only at extremely low frequencies for eg: via a decadal census. A number of papers have attempted predicting economic status at aggregated geographical levels such as district or neighborhood using Deep Learning on images with varying degrees of success. However the utility of such an approach at the household level remains open. In this study we utilize Deep Learning models on household images collected from four northeastern states in India to assess the feasibilty and ethics of household level income status prediction. We categorize households into classes based on income and then train a Swin Transformer model with cross-entropy loss and triplet loss to predict the socioeconomic class of the household. We then compare the prediction accuracy of our model with predictions using a simple list of household assets and predictions from a set of expert human annotators. We find that the use of Deep Learning on images does not lead to any substantial gains in prediction accuracy. Further, we note that human accuracy on this prediction tasks is low, raising questions on the information contained within the images. Our study raises important questions regarding the ethical implications of utilizing household images for predicting socioeconomic status. We explore these ethical implications, emphasizing the importance of a cautious and considerate approach in incorporating image-based techniques.

The Universal Immunization Programme in India has a mandate to fully vaccinate all of India's 27 million children born annually. The vaccination doses are recorded by frontline health workers on standardized paper-based Mother and Child Protection (MCP) cards, which are manually digitized by data entry operators, resulting in poor data quality, delays, and significant time and resources. In our article, we focus on Optical Character Recognition– (OCR) based automated digitization of MCP card images captured through a smartphone application developed by us. By utilizing a standardized template for the MCP cards, which is available
a priori
, we register the card images and perform OCR on the extracted region of interest (ROIs). Since the cards with curvature or torn edges had poor ROIs, we built a global–local alignment technique that first approximates the ROI using global homography and then refines using a local homography resulting in improved accuracy. Our pipeline gives a character level accuracy of 98.73% on our dataset against 75.02% by Google Cloud Vision and 79.26% by Azure OCR. We also describe our field testing experience, where the digitized MCP card images were used to provide useful features on the smartphone application for health workers to conduct vaccination sessions.