Skip to content

robinnarsinghranabhat/Data-Efficient-Cropland-Segmentation-with-Pseudo-labels

Repository files navigation

Introduction

Computer Vision Project for USD class Computer Vision CSC 752 -UT1

The project details can be found at : https://github.com/robinnarsinghranabhat/Data-Efficient-Cropland-Segmentation-with-Pseudo-labels/blob/main/CV_Report.pdf

Downloading the Sentinel-2 Images

You need to login to Google Earth Engine Explorer, and use the JS files in : ./gee/

  1. ee.js prepares the Export Files to be downloaded. We have two 27-channel image files on a provided region. Note : For a larger region, multiple files are downloaded into google drive which must be merged.

  2. You can select the region of your choice, or use the coordinates used in ./gee/final_roi.js

Notebooks

Refer

  • segmentation_v1.ipynb : train the baseline model
  • segmentation_unet_dic.ipynb : train Unet model with dice loss
  • segmentation_unet_bce.ipynb : train Unet model with Binary cross entropy loss
  • inference.ipynb : Compares different trained models againts each other

Results

Generating False Labels

alt text

Performance of different variants

alt text

IOU Score comparisons

alt text

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published