It is sooo exciting writing this from Stellenbosch, South Africa.
I’m here for the 2nd Deep Indaba, the brainchild of many including Shakir Mohamed (DeepMind), Moustapha Cisse (Facebook) and others. The atmosphere is incredible, the venue, fantastic.
I feel so much is converging now. Towards the end of the FDL 3.0 programme I visited the two FDL Europe teams, working in the University of Oxford, for Frontier Development Lab, Europe, the European arm of NASA FDL, led by the European Space Agency who are using artificial intelligence (AI) to enable better mapping and detection of informal settlements, defined as housing areas with no formal infrastructure ie slums. Team 1 included data on on the Kibera slum, Nairobi and last night I drove past the kilometres and kilometres of shanty towns outside Cape Town.
FDL EU team results have already improved identification from satellite images, based on the “hot tin roof” concept. Multi-spectral imaging allows up to 13 frequency bands to be detected. Metals like tin show up incredibly well and have proven to be an excellent proxy for these settlements, versus concrete or tile. Material differentiation is possible within at least 10 bands.
The team used open-source data from Afrobarometer and was assisted by UNICEF.
After final presentations in London, last month, the team submitted their work to Deep Indaba which will feature it on the website. They also submitted to AAAI and will present at a large event with ESA/ESRIN in Italy, November 2018.
This software will enable NGO’s to more accurately map and manage these settlements but also allow much faster and accurate disaster response. FDL EU Team 2 are working on just this.
We discussed the fact that residents from Kibera are – today – also being trained in data input and IT, by the wonderful Samasource. They also help train data specialists for labelling datasets, to enable teams and researchers around the globe to train AI too!
Another route into AI, as well as via Deep Indaba, is attending conferences. The excellent ‘Black in AI’ workshop is currently calling for submissions, offering grant applications to attend or to present at their NIPS 18 workshop in Montreal, Canada this December, apply here https://blackinai.github.io/
One potential AI coder-in-the-making is Peter, the oldest son of our late Kenyan liaison, Irene RIP. She would have been very proud that Peter was able to confirm his registration at the new Nachu Technical Training Institute today. Best Wishes to you, Peter. Study hard!