Code implementation for the lake detection contest in GISCUP 2023.
The submitted gpkg file for evaluation is located at ./GPKG/lake_polygons_test.gpkg.
- Ubuntu 20.04 LTS with Python 3.7.7
pip install -r requirements.txt- Suggested running environment: a virtual machine with an 8-core Intel Xeon CPU, 128GB RAM and an NVIDIA Tesla A100 GPU (40GB VRAM).
- Install the required packages by following the above `Requirements' section.
- Copy all datasets (download here) to
./data/. - [Option 1] Train the model from the origin, and evaluate it on the test dataset. The output gpkg for test regions is dumped to
./data/lake_polygons_test1.gpkg.
python main.py --dumpfile_uniqueid 1- [Option 2] Load the checkpoint of the pre-trained model and evaluate it on the test dataset, where the checkpoint file can be download from here. Move the downloaded pt file to
./exp/snapshot/. The output gpkg for test regions is dumped to./data/lake_polygons_test2.gpkg.
python main.py --load_checkpoint --dumpfile_uniqueid 2Email changyanchuan@gmail.com if you have any queries.
The source code for the site is under the MIT license, which you can find in the LICENSE.txt file.