Detectron2 on pytorch seems promising.

https://colab.research.google.com/drive ... i4hmGYk1dL

To allow local domain:-

ignoreip = 192.168.1.0/24 (that is to say 192.168.1.1-255)

also:

bantime = -1 (no expiry, hence permanent until restart)

Places where the domain name clashes:-

/etc/apache2/sites-enabled/default-ssl.conf
- ServerName in the <VirtualHost *:80> section

/etc/apache2/sites-enabled/000-default.conf
- ServerName and ServerAlias in the <VirtualHost *:80> section

The points of concern are:- /etc/apache2/sites-available/000-default.conf - add Listen 8054 (or whichever port you will be using) /etc/apache2/sites-available/app.conf (whatever your app happens to be) - <VirtualHost: *:8054> (or whichever port you will be using) /etc/apache2/sites-available/default...

For polygons (Choropleth style) in scattermapbox, you will need to define an array to submit as layers. This arrange will consist of one layer per polygon defined according to the following:- https://plot.ly/python/filled-area-on-mapbox/ In the case of geometries retrieved from onemapsg, these need ...

Choropleth management and configuration:-

https://plot.ly/~empet/14692/mapbox-cho ... th-plotly/#/

https://mothership.sg/2016/10/nifty-map ... h-is-no-1/

Note that density doesn't not in itself determine high or low number of people in the overall area. Bedok, for instance has lower density but higher population.

Looks like the Transformer architecture is better than the LSTM which is better, in turn, than RNNs, for handling time dependencies.