Search

Your search keyword '"Training data"' showing total 2 results
Start Over Descriptor "Training data" Language croatian
2 results on '"Training data"'

1. Evaluating training data for crop type classifıcation using support vector machine and random forests

Author

Mustafa Ustuner and Fusun Balik Sanli

Subjects
training data, crop type classification, support vector machines, random forests, machine learning, Geodesy, QB275-343
Abstract

This study evaluated the effectiveness of three different training datasets for crop type classification using both support vector machines (SVMs) and random forests (RFs). In supervised classification, one of the main facing challanges is to define the training set for the full representation of land use/cover classes. The adaptation of traning data, with the implemented classifier and its characteristics (purity, size and distribution of sample pixels), are of key importance in this context. The experimental results were compared in terms of the classification accuracy with 10-fold cross validation. Results suggest that higher classification accuracies were obtained by less number of training samples. Furthermore, it is highlighted that both methods (SVMs and RFs) are proven to be the effective and powerful classifiers for crop type classification.

Published
2017
Full Text
View/download PDF

2. Frontalizacija lica temeljena na dubokim generativnim modelima

Author

Mesec, Patrik and Jović, Alan

Subjects
model FFWM, identifikacija lica, deep generative models, training data, frontalizacija lica, FFWM model, TECHNICAL SCIENCES. Computing, TEHNIČKE ZNANOSTI. Računarstvo, duboki generativni modeli, podaci za učenje, face identification, face frontalization
Abstract

Suvremeni modeli identifikacije lica imaju manju točnost na profilnim pozama lica, posebice u ekstremnim slučajevima. Glavni cilj rada bio je izrada modela za ostvarivanje prednjeg pogleda na lice (frontalizacija) s ciljem poboljšanja točnosti identifikacije lica. Najvažniji dio rada bio je sakupljanje kritične količine podataka pomoću kojih je učeno nekoliko dubokih generativnih modela. Najbolji naučeni model temelji se na arhitekturi modela FFWM. Učene su različite varijacije modela FFWM, te je prikazan utjecaj pojedinih komponenti modela. Točnost identifikacije lica mjerena je na više različitih modela za izlučivanje značajki lica na skupu za evaluaciju. Frontalizacijom lica ostvarena su značajna poboljšanja identifikacije lica na svim modelima. Osim kvantitativne evaluacije, dana je i vizualizacija ostvarenih rezultata frontalizacije na skupu za evaluaciju i na dodatnim primjercima iz različitih izvora. Modern face identification models have worse accuracy on the profile poses of the face, especially in extreme cases. The main goal of this thesis was to develop a model for the face frontalization with the aim of improving the accuracy of face identification. The most important part of the work was the collection of a critical amount of data by which several deep generative models were trained. The best trained model is based on the architecture of the FFWM model. Different variations of the FFWM model were trained and the influence of some model components was shown. The accuracy of face identification was measured on several different models for extracting facial features on validation set. Face frontalization led to a significant improvement of face identification on all models. In addition to quantitative evaluation, the face frontalization results achieved on the validation set and additional examples from various sources were also visualized.

Published
2022

Catalog

Books, media, physical & digital resources
See catalog results