![]() ![]() Python baseline/foundation_eval.py -model_path /path/to/saved/foundation/best_model.pth -in_csv /path/to/val/or/test.csv -save_fig_dir /path/to/output/foundation/eval_test/pngs -gpu 0 -model_name resnet34 pngs for visual inspection of predictions Python baseline/foundation_eval.py -model_path /path/to/saved/foundation/best_model.pth -in_csv /path/to/val/or/test.csv -save_preds_dir /path/to/output/foundation/eval_test -gpu 0 -model_name resnet34 Write prediction tiffs to be used for postprocessing and generating the submission. Python baseline/train_foundation_features.py -train_csv /path/to/train.csv -val_csv /path/to/val.csv -save_dir /path/to/save/directory/foundation -model_name resnet34 -lr 0.0001 -batch_size 2 -n_epochs 50 -gpu 0 Inference with Foundation Features Network Train/Validate Foundation Features Network These csvs are used by the dataloader during training. This will create a csv file with filepaths to training images and labels and csv file with filepaths to validation images. Python baseline/data_prep/generate_train_val_test_csvs.py -root_dir /path/to/spacenet8/aws/data/download -aoi_dirs Germany_Training_Public Louisiana-East_Training_Public -out_csv_basename sn8_data -val_percent 0.15 -out_dir /path/to/output/folder/for/train/val/csvs binary road mask (0 non-building, 1 building)Ĭreate a random train/val split to train the models on binary building mask (0 non-building, 1 building) Python baseline/data_prep/create_masks.py -root_dir /path/to/spacenet8/aws/data/download -aoi_dirs Germany_Training_Public Louisiana-East_Training_Publicįour masks are generated during this process: shpĬreate training and validation masks from the geojsons to use during training and validation. The following new files will be written to the AOI annotations directory: It will output a few additional files that are used in the subsequent step for creating image masks. The cleaning step here also catches geometry problems, makes a single commom schema, and moves roads and buildings to seperate geojsons/shps. Python baseline/data_prep/geojson_prep.py -root_dir /path/to/spacenet8/aws/data/download -aoi_dirs Germany_Training_Public Louisiana-East_Training_Public Nvidia-docker run -it -rm sn8/baseline:1.0 bashįollow these steps to prepare data for training and validation.Ĭlean the geojson labels and add speed values to the roads based on road type, number of lanes, and surface type. ![]() Nvidia-docker build -t sn8/baseline:1.0 /path/to/sn8_baseline/dockerĬreate and run the container (mount volumes to access your data, etc. The goal of SpaceNet 8 is to leverage the existing repository of datasets and algorithms from SpaceNet Challenges 1-7 ( ) and apply them to a real-world disaster response scenario, expanding to multiclass feature extraction and characterization for flooded roads and buildings and predicting road speed. To help address this need, the SpaceNet 8 Flood Detection Challenge will focus on infrastructure and flood mapping related to hurricanes and heavy rains that cause route obstructions and significant damage. ![]() As these events become more frequent and severe, there is an increasing need to rapidly develop maps and analyze the scale of destruction to better direct resources and first responders. Algorithmic Baseline for SpaceNet 8 Flood Detection ChallengeĮach year, natural disasters such as hurricanes, tornadoes, earthquakes and floods significantly damage infrastructure and result in loss of life, property and billions of dollars. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |