IMAGE SEGMENTATION ACCURACY DEPENDING ON THE DEPTH OF U-NET MODEL
DOI:
https://doi.org/10.17770/het2020.24.6755Keywords:
accuracy, image segmentation, machine learning, neural networkAbstract
The aim of this work is to obtain information about impact of the depth of U-Net architecture model into segmentation accuracy. Experiment was completed using dataset of DSM images. Neural networks were trained to recognize building locations. Experiment considered to decrease the number of U-Net filter blokes to measure impact on result accuracy.Downloads
References
Forecasting: Principles and Practice [Tiešsaite] Pieejams: https://otexts.com/fpp2/nnetar.html [Piekļuve 15.04.2020]
An Introduction to Machine Learning by Anmol Behl [Tiešsaite] Pieejams: https://becominghuman.ai/an-introduction-to-machine-learning-33a1b5d3a560 [Piekļuve 15.04.2020]
Neural Networks for Image Recognition: Methods, Best Practices, Applications, [Tiešsaite] Pieejams: https://missinglink.ai/guides/computer-vision/neural-networks-image-recognition-methods-best-practices-applications/ [Piekļuve 15.04.2020]
Understanding Semantic Segmentation with UNET Pieejams:https://towardsdatascience.com/understanding-semantic-segmentation-with-unet-6be4f42d4b47 [Piekļuve 17.04.2020]
U-Net: Convolutional Networks for Biomedical Image Segmentation by Olaf Ronneberger, Philipp Fischer, and Thomas Brox. Computer Science Department and BIOSS Centre for Biological Signalling Studies, University of Freiburg, Germany, 2015.
