Welcome to fastgs

Introduction

This library is currently in alpha, neither the functionality nor the API is stable

This library provides geospatial multi-spectral image support for fastai. FastAI already has extensive support for RGB images in the pipeline. I try to achieve feature parity for multi-spectral images with this library, specifically in the context of Sentinel 2 geospatial imaging.

Sample Notebooks

You can find demo usage of the library in these notebooks

1. working with the kaggle 38-cloud/95-cloud landsat dataset.
2. working on a segmentation problem with a Sentinel 2 dataset

These are boths works in progress and optimized to display the features of the library, rather than the best possible results. Even so, the “cloud 95” notebook is providing results comparable to other hiqh quality notebooks on the same dataset.

Install

pip install -Uqq fastgs

conda install -c restlessronin fastgs

How to use

The low-level functionality is wrapped into a class that loads sets of Sentinel 2 channels into a multi-spectral tensor (a TensorImageMS subclass of fastai TensorImage which itself is a subclass of the pytorch Tensor).

from fastgs.multispectral import *

The following code creates a class that can load 11 Sentinel 2 channels into a TensorImageMS.

from fastgs.test.io import * # defines read_multichan_files_as_tensor

sentinel2 = createSentinel2Descriptor()

snt_12 = MSData.from_all(
    sentinel2,
    ["B02","B03","B04","B05","B06","B07","B08","B8A","B11","B12","AOT"],
    [sentinel2.rgb_combo["natural_color"], ["B07","B06","B05"],["B12","B11","B8A"],["B08"]],
    get_channel_filenames,
    read_multichan_files
)

The second parameter is a list of 4 channel sets that are minimally required to visualize all the individual channels.

img_12 = snt_12.load_image(66)
img_12.show()

[<AxesSubplot:>, <AxesSubplot:>, <AxesSubplot:>, <AxesSubplot:>]

This lower level functionality is made available by higher level wrapper APIs for creating data loaders, adding augmentations, displaying batches, results and top losses.

Acknowledgements

This library is inspired by the following notebooks (and related works by the authors)

• Mauricio Cordeiro’s multi-spectral segmentation fastai pipeline
• Nick Wright’s multi-spectral classification notebook