This repo contains weights of Unet++ model with SE-ResNeXt101 encoder trained with Istanbul, Inria and Massachusetts datasets seperately. Trainings have been realized using PyTorch and segmentation models library (https://github.com/qubvel/segmentation_models.pytorch) We also provide an inference notebook to run prediction on GeoTiff images. This notebook also outputs prediction images as GeoTiff.
We have addeed more weights of different architectures trained with Istanbul dataset.
You can use the following links to download weights files:
- Unet++ trained with Istanbul Dataset: Download
- Unet++ trained with Inria Dataset: Download
- Unet++ trained with Massachusetts Dataset: Download
New Weights trained with Istanbul Dataset:
- Unet++ with InceptionResNetv2 encoder: Download
- Unet++ with EfficientNet-b6 encoder: Download
- UNet with SE-ResNeXt101 encoder: Download
- UNet with InceptionResNetv2 encoder: Download
- UNet with EfficientNet-b6 encoder: Download
- DeepLabv3+ with SE-ResNeXt101 encoder: Download
To run the notebook, following libraries are required:
- torch == 1.7.1
- segmentation-models-pytorch == 0.1.3
- scikit-image == 0.18.1
- GDAL == 3.2.1
- tifffile == 2021.2.1
Bakirman, T., Komurcu, I. & Sertel, E., (2022) Comparative analysis of deep learning-based building extraction methods with the new VHR Istanbul dataset, Experts Systems with Applications, vol. 202, 117346, https://doi.org/10.1016/j.eswa.2022.117346.