👁 OpenEDS Sparse Semantic Segmentation Challenge

Scores table

Submission File	Config file	Fold 0	Leaderbord Score
baseline.json	baseline.yml	0.93276	0.94215
baseline-tta.json	baseline.yml	0.93358	0.94286
baseline-ensemble-tta.json	baseline.yml	-	0.94434

Training

python run_train.py <CONFIG_FILE> 

Any config param could be overridden in a such way:

python run_train.py <CONFIG_FILE> --epochs=1000 --checkpoint.filename=<CHECKPOINT_FILE> --checkpoint.model=True

Create submission

python run_inference.py <CONFIG_FILE>  <CHECKPOINT_FILE> <OUTPUT_FOLDER>
python create_json_ss.py --root-folder <OUTPUT_FOLDER> --submission-json <SUBMISSION_JSON>

Weights

https://drive.google.com/drive/folders/1kxgu-PVptBqsSWg7T9qwsaWJa2QlENSY?usp=sharing