Detecting Illegal Small Scale Mines in Ghana using Deep Learning

Computer vision is becoming one of the focus areas in artificial intelligence (AI) to enable computers to see and perceive like humans. In a collaboration with the Royal Holloway University, we applied deep learning to locate small scale mines in Ghana using satellite imagery, scaled training using Kubernetes and investigated their impact on surrounding populations & environment.

Episode 6 of the re:publica 18 - Science & Technology podcast, hosted by Olivia Klose, titled "Detecting Illegal Small Scale Mines in Ghana using Deep Learning" was published on May 2, 2018 and runs 26 minutes.

May 2, 2018 ·26m · re:publica 18 - Science & Technology

Summary

Olivia Klose Computer Vision enables computers to obtain a high-level understanding from images and videos by automatically extracting, analysing and understanding useful information. With autonomous driving, visual failure detection or scene understanding, computer vision is becoming one of the focus areas in artificial intelligence (AI) to enable computers to see and perceive like humans. In this talk we will present our ongoing collaboration with the Royal Holloway - University of London on illegal small scale mines in Ghana. Illegal small-scale mining is a growing industry in many African, Asian and Latin American developing countries. Gold and other precious minerals are extracted in a low-tech, labour-intensive process linked to environmental damages, health hazards and social ills. Additionally, the process provides huge employment and income potential in poverty-stricken communities. Since these small mining operations are mostly illegal, there is virtually no data to analyse their exact impact. This project seeks to fill this void to enable better-informed policy decisions by relevant stakeholders. We built an image classification model in Keras and scaled the training of the model using Kubernetes on Azure. Once small scale mines were identified, we investigated the impact of those mines on surrounding environments and populations in Python. supported by BMZ

Episode Description