search

alok-ai-lab / pyDeepInsight

A python implementation of the DeepInsight methodology.
GNU General Public License v3.0
158 stars 47 forks source link

How to plot a 2D image space formed after applying t-SNE #21

Closed ezekny closed 2 years ago

ezekny commented 2 years ago

I have been implementing the squeezeNet and wanted to plot the 2D image of t-SNE and how to apply the Convex Hull for rotation. Kindly assist on the way forward on this. Thank you

kaboroevich commented 2 years ago

The pre-rotated t-SNE positions are not stored. Two possible options are

  1. Set plot=True of the pyDeepInsight.ImageTransformer.fit method as seen in README.md. This will return an image of the t-SNE reduced features (blue), convex full (red), and minimum bounding rectangle (green) prior to rotation. There is no way to modify this image.
  2. Plot the post-rotation reduced features that are stored in the private attribute ImageTransformer._xrot.