| dc.description.abstract | Land use and land cover (LULC) classification have been generated for a wide range
of applications such as urban planning, environmental monitoring, etc. To respond,
a number of classification techniques have been developed, one of the most notable
approaches is Convolutional Neural Network (CNN). By utilizing satellite images, we
save time on field surveys and reduce the surveying costs. This thesis deploys U-Net
architecture - a fully convolutional neural network - with transfer learning on
ResNet-34 to classify Sentinel-2 image at the city scale. Sentinel-2 image acquired
on 13th January, 2020 of Da Lat City is used to construct manual labels training
dataset for our deep learning model. Afterward this study mapped the land cover
and land use classification of 7 classes: background, water, trees, crops, shrub, built
area, and bare land. | en_US |
