Android application for detecting small objects of the same type in photographs

U ovom radu opisan je razvoj Android aplikacije za detektiranje malih istovrsnih objekata koji predstavljaju mrave na slikama. Detektiranje objekata se temelji na modelu čija je izrada opisana od pripreme podataka do samog treniranja modela koristeći TensorFlow te Google Colab. Razvijena su dva modela sa razlikama u skupu podataka za treniranje te su objašnjene njihove razlike. Prikazano je i sučelje aplikacije koja, od dva modela, implementira onaj efikasniji naučeni model za detekciju objekata. Također su opisane sve korištene tehnologije.This thesis describes development of an Android application fot detecting small objects of the same type that represents ants in photographs. Object detection is based on a model, the creation of which is described from data preparation to model training using TensorFlow and Google Colab. Two models with differences in the training dataset were developed and their differences were explained. The application interface is also shown, which, of the two models, implements the more efficient trained model for object detection. All the technologies used are also described