wchch1010 / caffe-cvmlp-polsar

Implementation of Complex-Valued Convolutional Multilayer Perception(CV-MLP) for classification of PolSAR image with pixels that are represented as complex value

Under DEV : caffe-cvmlp-polsar Caffe complex valued neural network for classification of PolSAR images

**This is a modified version of Caffe used to achive goals set for master thesis work

Task of this thesis is to implement and investigate the Complex-Valued Convolutional Multi-layer Perception(CV-MLP) in order to solve classiffcation problem of PolSAR Images.

In other words given PolSAR image pixels that are represented as complex-valued scattering vector our task is to automatically label regions on PolSAR image, for example: vegetation, buildings, roads and more by mapping each image pixel to land type. In order to train CV-MLP we're going to use ground truth data that was manually labeled by humans.

Dataset

Manually labeled image "Oberpfaffenhofen" of size $1390 * 6640$ with 5 classes(city, field, forest, grassland, street).\ Training dataset generated using 4000 randomly selected pixels for each labeled class by taking region of $15 \times 15$ pixels.\ Total 20000 entries. Testing dataset contains same amount of image patches from remaining pixels after random selection for training.

Net specification

TBD

License

This extension to the Caffe library is released under a creative commons license which allows for personal and research use only. For a commercial license please contact the authors. You can view a license summary here: http://creativecommons.org/licenses/by-nc/4.0/