Open Rub21 opened 10 months ago
Thanks for working through this & detailed documentation @yunica @Rub21 - I did some reading around compass angle & their representations with true north. Convert compass angles to cartesian coordinates is the right approach.
Do we need any more information to align the points with building footprints?
@Rub21 Can you fix the asset links?
@srmsoumya , I have just updated the images and files.
@srmsoumya, following our discussion about the heading angles for Mapillary images, we have been assessing the images for use in post-processing to match the predictions with the buildings, similar to our approach in the previous phases of the Housing Passport project.
The 'compass_angle' is an attribute calculated by Mapillary when 360-degree images are uploaded. According to Mapillary's documentation/GhatGPT, the 'compass_angle' represents as navigation angle in Mapillary. details : https://gist.github.com/Rub21/228f9b144ce1fe7ad1b1a83caf053bd3
To visualize what the navigation heading angle looks like, I created script to generate LineString extending from the start point to a specific distance. However, the heading direction doesn't seem to align with the road:
Uing chat GPT and documentation Junior, we figured out that the navigation angle could be converted to Cartesian angles for use. The Cartesian angle representation is shown below:
the cartesian angle make more sense in terms of roads alienation, we can also determine the angles to the left and right sides. Here's how it looks:
As per our methodology in previous phases of the Housing Passport project, we need the cartesian angles for the left an right side to do post processing.
Here the script that evaluates the angles. https://github.com/developmentseed/housing-passports-scripts/pull/5
lines_outputs_navi.geojson.zip
points_outputs_navi.geojson.zip
I would like you inputs around this , or also to do further investigation.
cc. @geohacker @yunica