URnano: Nanopore base-calling from a perspective of instance segmentation

docker pull yaozhong/keras_r_tf:1.8

nvidia-docker run -it --rm yaozhong/keras_r_tf:1.8 bash

# setting model parameters
## Network structure
MODEL="../experiment/model/URnet.model.parameters.json"
## Training data cache
TRAINDATA="../data/cache/train.h5"
## If no data cache is given, provide the signal and labels as used in Chiron.
TRAIN_IN="../data/chiron_data/train/"
## Currency strength statistics saving path
NORM_FILE_SAVE="../experiment/model/statistics/all_data_stats.pickle"

python train_urnet.py -cf $TRAINDATA -i $TRAIN_IN -mp $MODEL -l $LOSS -nID 3 -fSignal 10 -norm $NORM_FILE_SAVE -tag URnet.all

## Network structure
MODEL="../experiment/model/URnet.model.parameters.json"
LOSS="ce_dice_loss"
TESTDATA="../data/cache/test.h5"
## If no data cache is given, provide the signal and labels as used in Chiron.
TEST_IN="../data/chiron_data/test/"
## Currency strength statistics saving path
NORM_FILE_SAVE="../experiment/model/statistics/all_data_stats.pickle"

python test_urnet.py -tm plt -cf $TESTDATA -i $TEST_IN -mp $MODEL -l $LOSS -nID 3 -norm $NORM_FILE_SAVE -tag URnet.all

spiece="ecoli"
SIGNAL_FOLD="../data/chiron_data/paper_eval/unet_result/signals/ecoli/"
OUTPUT="../experiment/basecall/basecalling_clip/$spiece/"
MODEL="../experiment/model/Unet.model.parameters.json"
LOSS="ce_dice_loss"
NORM_FILE_SAVE="../experiment/model/statistics/all_data_stats.pickle"

python fast5_test_urnet.py -i $SIGNAL_FOLD -it signal -o $OUTPUT -mp $MODEL -loss $LOSS -nID 3 -norm $NORM_FILE_SAVE -tag URnet.all