Search

We present a novel method for retrieval of chess positions similar to a given query position from a collection of archived chess games. Our approach considers not only the static similarity due to the arrangement of the chess pieces, but also the dynamic similarity based on the recognition of chess motifs and dynamic, tactical aspects of position similarity. We use information retrieval techniques to enable efficient approximate searches by encoding chess tactical problems as text documents. In addition, we designed and implemented a procedure for automatic generation of tactical puzzles from a collection of chess games. We have shown experimentally how important the inclusion of both static and dynamic features is for successful detection of similar chess motifs. In another experiment, the program was able to quickly traverse a large database of positions to identify similar tactical problems. A chess expert found the resulting program useful for automatically generating instructive examples for chess training. Predstavimo novo metodo za iskanje šahovskih pozicij, podobnih določeni poizvedbi v zbirki šahovskih partij. Naš pristop poleg statične podobnosti zaradi podobne postavitve figur upošteva tudi dinamično podobnost na podlagi prepoznavanja šahovskih motivov in dinamičnih, taktičnih vidikov podobnosti pozicij. Uporaba metod informacijskega poizvedovanja z zapisovanjem šahovskih taktičnih problemov v tekstovni obliki nam omogoči učinkovito poizvedovanje po obstoječi bazi pozicij. Predstavimo tudi postopek za avtomatsko generiranje taktičnih problemov iz zbirke šahovskih partij. S prvim eksperimentom smo pokazali pomembnost upoštevanja tako statičnih kot dinamičnih lastnosti pozicije za uspešno prepoznavanje podobnih šahovskih motivov. Z drugim eksperimentom smo pokazali učinkovitost programa za poizvedovanje po večji bazi pozicij. Šahovski ekspert je program prepoznal kot uporaben za avtomatsko generiranje poučnih primerov za šahovski trening.

Pri diplomskem delu, kjer se osredotočam na požare v naravnem okolju, sem raziskoval kako lahko tak požar, kar se da učinkovito pogasimo. Predstavil sem kdaj se določen požar sploh dojema kot požar v naravnem okolju, kakšne vrste le teh poznamo, vzroke za njihov nastanek ter osnove gašenja takih požarov. Različni taktični postopki, v katere so te osnove gašenja implementirane, in katere sem v delu tudi povzel, pa zagotavljajo učinkovito gašenje, ki je en izmed bolj pomembnih delov zagotavljanja varnosti in varovanja zdravja gasilcev. Raziskoval sem opremljenost gasilskih enot ter kako in na kakšen način se to opremo uporablja pri gašenju. S pomočjo strokovne literature sem opisal možnosti gašenja, ki so na voljo tako pri velikih in intenzivnih, kakor tudi pri manjših in manj intenzivnih požarih. Spoznal sem, da je gašenje zapleten, nevaren proces, ki zahteva obilo znanja ter izkušenj, da se izpelje hitro in kar se da varno. In my dissertation, where I focus on wildfires, I researched how such a fire can be extinguished as effectively as possible. I presented when a certain fire is perceived as a wildfire, what types of fires we know, the causes of their occurrence and the basics of extinguishing such fires. The various tactical procedures in which these basics of firefighting are implemented, and which I have also summarized in my work, ensure effective firefighting, which is one of the more important parts of ensuring the safety and health of firefighters. I researched the equipment of fire brigades and how and in what way this equipment is used in firefighting. With the help of literature, I have described the possibilities of firefighting large and intensive, as well as smaller and less intense fires. I have learned that extinguishing is a complex, dangerous process that requires a wealth of knowledge and experience to be carried out quickly and as safely as possible.

Za dobro delovanje potrebujejo globoke nevronske mreže veliko podatkov. V primeru biometrične modalnosti uhljev - največje anotirane baze slik uhljev v nekontroliranem okolju zajemajo nekaj tisoč slik, kar je premalo za globoko učenje in razpoznavo. Ta problem skušamo rešiti z uporabo generativnih nevronskih mrež za obogatitev baze. Implementiramo dva tipa generativnih nevronskih mrež: generativno mrežo in variacijski avtokodirnik. Obe mreži naučimo s pomočjo slik iz obstoječe baze in z vsako generiramo množico umetnih podatkov (slike uhljev). Z vsako od teh množic nato učimo mreže za razpoznavo in primerjamo rezultate. Kljub uporabi umetno generiranih slik, ne uspemo doseči visoke stopnje razpoznave na bazi AWE-v1, vseeno pa so opazne izboljšave v primerjavi z rezultati učenja razpoznave brez umetno generiranih slik. Deep neural networks require large amounts of data to perform well. In the case of the biometrical modality of the human ear, the largest annotated databases of images of ears in an uncontrolled environment consist of a few thousand images, which is insufficient for recognition using deep learning. We try to solve this problem using generative neural networks for data augmentation. We implement two types of generative neural networks: a generative network and a variational autoencoder. We train both networks on images from the existing database and then use them to generate a new set of artificial data (images of ears) with each. We then use each of these datasets to train neural networks for recognition and compare the results. Even using artificially generated images, we do not manage to achieve a high recognition rate on the AWE-v1 ear database. Despite that, there is a noticeable improvement compared to results of training for recognition without using generated data.

Deep neural networks require large amounts of data to perform well. In the case of the biometrical modality of the human ear, the largest annotated databases of images of ears in an uncontrolled environment consist of a few thousand images, which is insufficient for recognition using deep learning. We try to solve this problem using generative neural networks for data augmentation. We implement two types of generative neural networks: a generative network and a variational autoencoder. We train both networks on images from the existing database and then use them to generate a new set of artificial data (images of ears) with each. We then use each of these datasets to train neural networks for recognition and compare the results. Even using artificially generated images, we do not manage to achieve a high recognition rate on the AWE-v1 ear database. Despite that, there is a noticeable improvement compared to results of training for recognition without using generated data.