The objective of this thesis is to perform a comparative analysis of different convolutional neural network (CNN) models for the semantic segmentation of segetal flora on unmanned aerial vehicle (UAV) images. The study uses a dataset of segetal flora RGB images collected at different flight heights. The expected outcome includes a thorough analysis of the performance of at least three CNN models in terms of segmentation accuracy, computational efficiency, and transferability across multiple UAV flight heights. The research will provide insight into the strengths and limitations of each model, and provide valuable recommendations for their application in biodiversity monitoring.