Proceedings of the Conference on Modeling and Analysis of Complex Systems and Processes 2020 (MACSPro 2020)
In this paper, we study automatic recognition and counting of vehicles in the wild. For this problem, we tested several object detection models for car type recognition among five classes: Bicycle, Bus, Car, Motorcycle, Truck, Van. We extend existing dataset in order to balance classes and achieve classification quality for detected cars with 92% mAP
The paper describes a system that extracts facts and opinions from documentary texts to create a domain ontology of a controversial topic for Chernobyl disaster. The pipeline of the system is based on RNNbased NER module, which was tested on an annotated text corpora.
We consider the task of content-based video retrieval (CBVR) given a query video, which is expected to match if it is a distorted short subsequence of a reference video from a database. In this paper, we present a CBVR system architecture that is both robust and scalable. We use a modified rHash frame fingerprint generation method. It is both, extremely robust to distortions and fast to compute. We utilize the Faiss library, developed by Facebook Research, to index fingerprint binary vectors. The VCDB dataset is used for benchmarking.
Basel II and III allow banks to use own default statistics when estimating regulatory parameters (risk-weights) for the capital adequacy ratio purpose. Bank inputs own risk estimates into the Vasicek model. It yields a distribution of credit losses. Regulator then requires a bank to take 99.9% quantile of such a distribution as a risk-measure (a risk-weight). When saying regulator we mean any Central Bank (including Bank of Russia, but no limited to it) that allow local banks to run the described approach. Having being criticized for excessive conservatism, we reveal that it still underestimates credit risk. This comes from the newly discovered fact that the default correlation may tend to co-depend with the systemic factor (for instance, with the GDP growth rate), albeit originally such co-dependence was not considered. We use the US statistics on total loans defaults for 1985-2019 to evidence the finding. The credit loss underestimation thus at least exceeds by 11% the loss estimates using the maximum (100%) correlation with the systemic factor.