Deep Learning v analýze obrazu

Abstract

Theme of this master's thesis is deep learning in image analysis. The thesis describes principles of (artificial) neural networks and describes several types of neural networks followed by familiarization and practical use for detection of objects in image. Convolution neural networks are chosen to demonstrate detection of free and occupied parking places. This work is based on object-oriented and scripting programming language Python and framework Caffe. The framework Caffe also works on CUDA platform.