README.md

metadata

license: apache-2.0
language:
  - en
tags:
  - Pytorch
  - mmsegmentation
  - segmentation
  - Flood mapping
  - Sentinel-2
  - Geospatial
  - Foundation model
metrics:
  - accuracy
  - IoU

Model and Inputs

The pretrained Prithvi-100m parameter model is finetuned to segment the extend of floods on Sentinel-2 images from the Sen1Floods11 dataset.

The dataset consists of 446 labeled 512x512 chips that span all 14 biomes, 357 ecoregions, and 6 continents of the world across 11 flood events. The benchmark associated to Sen1Floods11 provides results for fully convolutional neural networks trained in various input/labeled data setups, considering Sentinel-1 and Sentinel-2 imagery.

We extract the following bands for flood mapping:

1. Blue
2. Green
3. Red
4. Narrow NIR
5. SWIR 1
6. SWIR 2

Labels represent no water (class 0), water/flood (class 1), and no data/clouds (class 2).

The Prithvi-100m model was initially pretrained using a sequence length of 3 timesteps. Based on the characteristics of this benchmark dataset, we focus on single-timestamp segmentation here. This demonstrates that our model does not require multiple timestamps during finetuning.

Code

Code for Finetuning is available through github

Configuration used for finetuning is available through this config.

Results

The experiment by running the mmseg stack for 80 epochs using the above config led to the following result:

Classes	IoU	Acc
No water	96.90%	98.11%
Water/Flood	80.46%	90.54%

aAcc	mIoU	mAcc
97.25%	88.68%	94.37%

Inference

The github repo includes an inference script that allows to run the flood mapping model for inference on Sentinel-2 images. These input have to be geotiff format, including 6 bands for a single time-step described above (Blue, Green, Red, Narrow NIR, SWIR, SWIR 2) in order. There is also a demo that leverages the same code here.