Self-supervised learning of depth estimation for uncalibrated cameras
| Autor: | Pašalić, Ante |
|---|---|
| Přispěvatelé: | Marković, Ivan |
| Jazyk: | chorvatština |
| Rok vydání: | 2022 |
| Předmět: |
umjetna inteligencija
umjetne neuronske mreže kalibracija kamere procjena dubine umjetne neuronske mreže TECHNICAL SCIENCES. Computing TEHNIČKE ZNANOSTI. Računarstvo umjetna inteligencija kalibracija kamere procjena dubine depth estimation artificial intelligence camera calibration artificial neural networks |
| Popis: | U ovom Završnom radu opisani su koncepti umjetnih neuronskih mreža te konvo- lucijskih neuronskih mreža. Opisan je naˇcin treniranja neuronskih mreža, te prikladnih skupova podataka za treniranje tih mreža. Osim toga, navedeni su temeljni koncepti kalibracije kamere. Tako ̄der je opisana struktura projekta Monodepth2 za procjenu dubine slike, koja je zatim nadogra ̄dena koriste ́ci zasebnu konvolucijsku neuronsku mrežu za kalibraciju kamere. In this BSc Thesis, the concepts of artificial neural networks and convolutional neural networks are described. Aside from that, the method of training neural networks and suitable datasets for training those networks are described. In addition, the basic concepts of camera calibration are outlined. Also described is the structure of the Monodepth2 project for image depth estimation, which was then upgraded using an additional convolutional neural network for camera calibration. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|