OSM Mapping with AI from Facebook

Facebook and HOT OSM have come together to use machine learning to make maps better. It looks like a mutually beneficial collaboration that will help OSM

“AI is supercharging the creation of maps around the world” is a very interesting development in OSM mapping. It perhaps will reduce the number of hours required to map areas by at least a couple of orders of magnitude.

Automatic feature extraction is nothing new of course. People have been doing it for decades. I should know, I spent two years mastering (supposedly) on Remote Sensing, the domain that deals with capturing information about the earth using satellites and extracting useful information from the captured data. The traditional remote sensing workflow uses techniques like Supervised Classification and Unsupervised Classification for feature extraction. The beauty of machine learning algorithms is that models can be built which can merge these two classification techniques and automate the process.

OSM community AFAIK stayed away from automated data generation mostly because the work involved in cleaning up bad data is more difficult and tedious than in creating new data. It looks like the collaboration from HOT OSM and the development of RapID (enhanced ID), the process put in place ensures that only quality data goes into the system preventing the issues of bad data from happening.

While it is all colourful for the OSM community, the question with profit seekers like Facebook is always what’s the catch? I tend to think there is no catch for the OSM community and that FB has decided to invest some money in the OSM so that they can use the fruits of community work for their benefits. It looks like a mutually beneficial arrangement. The community gets the resources from FB to create a better dataset and FB gets the data for its usage. I am sure FB will extract more value out of their investment in the long run and reap more benefits than the community. They will obviously have their own data layers on top of the OSM data, they will overlay all sorts of tracking data and enhance their capabilities as a surveillance machine. But all that is left to them and their corporate interests.

For now this is a good thing and doesn’t seem to do any harm to OSM.

Updating the Wikipedia Tamilnadu Literacy Map

On 16th October 2011, I have uploaded a map of Tamilnadu District wise Literacy levels to Wikipedia. It was used in the article about Tamilnadu for a long time, then moved to the Education in Tamilnadu article when a separate article was created. But the map was not in line with the Wikipedia Map Conventions. So, took some time this week and updated the map.

Updated Version

Tamil_Nadu_Literacy_Map_2011

Older version

2018122413022521Tamil_Nadu_Literacy_Map_2011

A Map of the Chetpet Lake and Eco Park

Chetpet Lake has been developed into a nice waterfront park for walking a few years back. It is maintained diligently with water level balanced between the two parts of the lake depending on the availability, grass moved, plants cared for and the walkways washed off the bird droppings everyday morning. It opens for walkers as early as 4.30 in the morning every day. It is one of the places in Chennai, that I have walked into and really felt peaceful.

It has boating, children’s play area, angling points, 3D theatre, multilevel car park, and a food court. It is well connected by public transport. It has a bus stop, a railway station, and a metro station right outside its walls. But, it didn’t have a map. So, I downloaded a PDF from OpenStreetMap and created one which is now used in Chetput Lake Wikipedia article.

ChetpetLake