I have a trained network that is giving me features of 2048 dimensions. I want to visualize them using t-sne plot through a Tensorboard embedding projector. Each of the data points belongs to some 10 different labels so this distinction I can do by specifying the different colors to different labels. However, the catch is, within 1 label, these features belong to 2 categories, one is the RGB image feature and the other is the infrared image feature. So now I need to show some marker like 'x' or 'o' which is available in Scatter plot different for RGB and infrared features. However, I couldn't find this feature to specify marker anywhere in Embedding Projector documentation. Currently, I am using Pytorch's embedding projector using Summary writer but I don't mind switching to Tensorflow if it has such functionality. Kindly help me if anybody knows if this can be done in a Tensorboard embedding projector. Thanks in advance

0

There are 0 best solutions below