deletions | additions       diff --git a/figures/nutshell/caption.tex b/figures/nutshell/caption.tex  index 90701ac..a5d6d60 100644  --- a/figures/nutshell/caption.tex  +++ b/figures/nutshell/caption.tex 
...
\label{fig:highleveloverview}  The figure illustrates the proposed system from a high-level perspective. The current camera image of the UAV is used to extract an image feature vector. The feature vector is fed into the machine learning model that uses a $k$-Nearest Neighbors algorithm to output $x,y$-position estimates. These estimates are forwarded to the particle filter, which filters position estimates over time and outputs a final position estimate (green circle). The expected loss shows regions in the map where the algorithm accuracy is expected to be lower. The average expected loss assigns a fitness value to a given map.