search

srimannarayanabaratam / land_cover_classification_unet

This repository contains code for implementing multi class semantic segmentation (specifically applied to satellite image classification) with PyTorch implementation of UNet.
https://baratam-tarunkumar.medium.com/land-cover-classification-with-u-net-aa618ea64a1b
44 stars 22 forks source link

Using Model for Different Data Shape #3

Closed gmvincent closed 3 years ago

gmvincent commented 3 years ago

I was wondering if this model/algorithm can be adapted to use a 9-channel hyperspectral image instead of an RGB image? Or can it be adapted to use an image with a 4D shaped tensor [time, height, width, channel (9)]?

srimannarayanabaratam commented 3 years ago

Hi there!

I am sure you can try more than 3 input cannels with UNet architecture.

However, for 4D tensors with time you might have to look at 3D convolutions. An inflated version of 2D conv. Time dimension is unlike geometrical dimension.

On Mon, Aug 2, 2021, 5:19 AM gmvincent @.***> wrote:

I was wondering if this model/algorithm can be adapted to use a 9-channel hyperspectral image instead of an RGB image? Or can it be adapted to use an image with a 4D shaped tensor [time, height, width, channel (9)]?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/TarunKumar1995-glitch/land_cover_classification_unet/issues/3, or unsubscribe https://github.com/notifications/unsubscribe-auth/AS6XO6BHPNLXMDA225H3JDDT2XMPFANCNFSM5BLUM4WQ .

gmvincent commented 3 years ago

@TarunKumar1995-glitch, thank you!