Generátor a detektor modelů zbraní

This report contains description of method for detection, classification of type (length) and orientation (rotation) of weapon. Individual weapon types, machine learning methods both classical and neural networks approaches are described. In the case of neural networks individual layers and used architectures are described. Detectors of objects are also described. For each part (detection, type and orientation) several approaches are designed. These are tested on available data. For rotation dataset is created using automatic generation. In conclusion results are discussed in case of detection 97.8 % in length 84.14 % and in rotation 75.14 %.