qAp/omdena_engie

omdena_engie

Work done for the Omdena-Engie project Increasing Solar Energy Adoption Through Roof Detection

This work consists of Jupyter and Colab Notebooks, a summary of which can be found at my blog post Super-resolving satellite images using ESRGAN.

Brief description of each notebook in this repository:

• 01_image_preprocessing.ipynb
  Crop large satellite image into tiles using geopandas, supermercado, rio_cogeo, and rio_tiler.

• 01a_image_preprocessing.ipynb
  Convert all images to RGB images.

• 02_pretrained_ISR_inference.ipynb
  Apply super-resolution to satellite images using pre-trained ESRGAN available at idealo/image-super-resolution.

• 02a_ISR_training.ipynb
  Training the ESRGAN implementation at idealo/image-super-resolution.

• 03_sr_satellite_datasets.ipynb
  Display and compare satellite images of various resolutions: SpacenetV2 (0.3 m), New Zealand Land Services (0.1 m), xView2 (0.3 m), and Solar-AI's (~ 0.7 m).

• 04_pretrained_ESRGAN-pytorch_inference.ipynb
  Applying super-resolution to satellite images using pre-trained ESRGAN available at wonbeomjang/ESRGAN-pytorch.

• 05_generate_LR_images_colab.ipynb
  Generate high resolution and low resolution image pairs for training ESRGAN.

• 06_ESRGAN-pytorch_training_colab.ipynb
  Training ESRGAN-pytorch on Colab on satellite image data.

• 07_esrgan_inference_colab.ipynb
  Inference with ESRGAN trained on satellite images.

• 08_building_segmentation_inference.ipynb
  Using pre-trained building segmentation model at kbrodt/building-segmentation-disaster-resilience on Solar-AI's images.