8000 GitHub - kevindean/LineProfileDeconvolution: uses line profiles to semantically segment an image in 1 Dimension
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

kevindean/LineProfileDeconvolution

Repository files navigation

LineProfileDeconvolution

uses line profiles to semantically segment an image in 1 Dimension

Note -> this is an initial release. The network requires more data (which will be coming soon) in order to get closer to the truth; as well as not overfit.

Data Acquisition from Line Profiles from the vertical direction through a volume Data Acquisition

Transcribe the data down to 1 Dimension Transcribe to Neural Net Input Data

Train the Network (see how to adjust the network structure / dataset to get the best results; the network is set up for semantic segmentation, represented as a unet-style architecture; utilized Keras for quick plug and play) Model Architecture

Network Prediction of Line Profile in 1 Dimension Network Prediction of Line Profile in 1 Dimension

Using a Point Locator, keep track of the indices and map the predictions back to a 3 Dimensional Volume Mapped back into a 3 Dimensional image as ---> Input, Prediction, Truth

2nd Note -> Analysis module will be coming soon.

About

uses line profiles to semantically segment an image in 1 Dimension

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

0