Multi-directional and Multi-constraint Learning Network for Aerial Imagery Semantic Segmentation

Folder Structure

Prepare the following folders to organize this repo:

├── MMLN (code)
├── pretrain_weights (save the pretrained weights like vit, swin, etc)
├── model_weights (save the model weights)
├── fig_results (save the masks predicted by models)
├── lightning_logs (CSV format training logs)
├── data
│   ├── LoveDA
│   ├── uavid
│   ├── vaihingen
│   ├── potsdam 

Install

Open the folder airs using Linux Terminal and create python environment:

conda create -n airs python=3.8
conda activate airs

conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge
pip install -r GeoSeg/requirements.txt

Pretrained Weights

Quark Netdisk : 7zbh

Data Preprocessing

Quark Netdisk : aGvz

Training

python MMLN/train_supervision.py -c MMLN/config/uavid/***.py

Use different config to train different models.

Validation

For example:

python MMLN/loveda_test.py -c MMLN/config/loveda/***.py -o fig_results/loveda/*** --rgb --val -t 'd4'

Testing

LoveDA

python MMLN/loveda_test.py -c MMLN/config/loveda/***.py -o fig_results/loveda/*** -t 'd4'

UAVid

python MMLN/inference_uavid.py \
-i 'data/uavid/uavid_test' \
-c MMLN/config/uavid/***.py \
-o fig_results/uavid/*** \
-t 'lr' -ph 1152 -pw 1024 -b 2 -d "uavid"

Inference on huge remote sensing image

python MMLN/inference_huge_image.py \
-i data/vaihingen/test_images \
-c GeoSeg/config/vaihingen/***.py \
-o fig_results/vaihingen/*** \
-t 'lr' -ph 512 -pw 512 -b 2 -d "pv"

Reproduction Results

Due to some random operations in the training stage, reproduced results (run once) are slightly different from the reported in paper.

Acknowledgement

Thanks to the GeoSeg framework for the help of the above code implementation of this work, and to the hardware support provided by the Advanced Computing Center of China Three Gorges University.