Najčešće korišćen izvor za dobavljanje podataka o poljoprivrednim granicama je katastar. Međutim, proces ekstrakcije katastarskih podataka je i dalje značajno manuelan proces, što neizbežno dovodi do povećane verovatnoće za grešku u merenju, kao i pove­ćanih troškova. Jedan od mogućih načina prevazilaženja spomenutih problema je automatizacija procesa delineacije granica parcela korišćenjem satelitskih slika i metoda dubokog učenja. U ovom radu predstavljena je konvolu­tivna neuronska mreža za detekciju (tj. delineaciju) poljoprivrednih granica, gde se kao ulaz modela koriste multi-spektralne satelitske slike. Istrenirani model uspešno prepoznaje granice poljoprivrednih parcela i ostvaruje rezultate bolje od autora originalnog skupa podataka. Iako su ostvareni zadovoljavajući rezultati, predloženi su koraci za poboljšanje predloženog rešenja koji bi mogli dovesti do ostvarivanja robusnijih rezultata.

The most common source of field boundary data is cadaster. Unfortunately, the process of extracting cadastral information has remained highly manual, which inevitably leads to proneness to errors and high costs. One of the possible ways to improve generating and updating cadastral information is automating the delineation of boundaries through satellite imagery and deep learningbased methods. This paper presents a CNN-based architecture for delineating agricultural boundaries, where multispectral satellite images are used as an input of the model. The trained model performs accurate boundary detection and achieves better results than the authors of the original dataset. Even though satisfactory results were achieved, we proposed steps to improve the proposed solution, which could increase robustness.

In this paper, we present an initial study of possibilities of applying Artificial Intelligence (AI) and computer vision-based approaches aimed to improve and alleviate the process of conducting knowledge assessment over the Microsoft Teams platform. We did that by developing a deep-neural-network-based system which is able to locate faces and predict emotions based on students' facial expressions. The system was evaluated on videos recorded during an online assessment of the ability of students to train and deploy deep learning solutions using Python and TensorFlow. We present results of this evaluation and show that, although the accuracy of our algorithm is limited at frame level, as we optimized for computational performance, it provides sufficient information to identify key changes on students' behavior, which should be brought to the teachers' attention.