🐙

Elvis Saravia omarsar

🐙

Machine Learning & NLP Research • PhD • Building @dair-ai • Previously: Meta AI, Elastic

omarsar / mt_model.py

Created October 15, 2018 09:15

	model = mt_models.Unet(drop_rate=0.4, bn_momentum=0.1)
	model.cuda()
	num_epochs = 10
	initial_lr = 0.001
	optimizer = optim.Adam(model.parameters(), lr=initial_lr)
	scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer, num_epochs)

omarsar / mt_dataloaders.py

Created October 15, 2018 09:09

	train_loader = DataLoader(gmdataset_train, batch_size=16,
	shuffle=True, pin_memory=True,
	collate_fn=mt_datasets.mt_collate,
	num_workers=1)

	val_loader = DataLoader(gmdataset_val, batch_size=16,
	shuffle=True, pin_memory=True,
	collate_fn=mt_datasets.mt_collate,
	num_workers=1)

omarsar / mt_split_data.py

Created October 15, 2018 09:08

	### training dataset with 80/20 split
	TRAIN_ROOT_DIR_GMCHALLENGE = "/gdrive/My Drive/DAIR RESOURCES/PyTorch/medical_imaging/train/"

	gmdataset_train = mt_datasets.SCGMChallenge2DTrain(root_dir=TRAIN_ROOT_DIR_GMCHALLENGE,
	subj_ids=range(1, 9),
	transform=train_transform,
	slice_filter_fn=mt_filters.SliceFilter())

	gmdataset_val = mt_datasets.SCGMChallenge2DTrain(root_dir=TRAIN_ROOT_DIR_GMCHALLENGE,
	subj_ids=range(9, 11),

omarsar / mt_transformations.py

Created October 15, 2018 09:04

	# training dataset
	train_transform = transforms.Compose([
	mt_transforms.Resample(0.25, 0.25),
	mt_transforms.CenterCrop2D((200, 200)),
	mt_transforms.ElasticTransform(alpha_range=(28.0, 30.0),
	sigma_range=(3.5, 4.0),
	p=0.3),
	mt_transforms.RandomAffine(degrees=4.6,
	scale=(0.98, 1.02),
	translate=(0.03, 0.03)),

omarsar / mt_prediction_th.py

Created October 15, 2018 09:03

	def threshold_predictions(predictions, thr=0.999):
	thresholded_preds = predictions[:]
	low_values_indices = thresholded_preds < thr
	thresholded_preds[low_values_indices] = 0
	low_values_indices = thresholded_preds >= thr
	thresholded_preds[low_values_indices] = 1
	return thresholded_preds

omarsar / mt_transform.py

Created October 15, 2018 08:58

	# transformer
	composed_transform = transforms.Compose([
	mt_transforms.Resample(0.25, 0.25),
	mt_transforms.CenterCrop2D((200, 200)),
	mt_transforms.ToTensor(),
	])

	# load data
	train_dataset = mt_datasets.SCGMChallenge2DTrain(root_dir=ROOT_DIR_GMCHALLENGE, transform=composed_transform)

omarsar / mt_explore.py

Created October 15, 2018 08:50

	ROOT_DIR_GMCHALLENGE = "/gdrive/My Drive/DAIR RESOURCES/PyTorch/medical_imaging/train/"
	mri_input_filename = os.path.join(ROOT_DIR_GMCHALLENGE,
	'site1-sc01-image.nii.gz')
	mri_gt_filename = os.path.join(ROOT_DIR_GMCHALLENGE,
	'site1-sc01-mask-r1.nii.gz')

	pair = mt_datasets.SegmentationPair2D(mri_input_filename, mri_gt_filename)
	slice_pair = pair.get_pair_slice(0)
	input_slice = slice_pair["input"]
	gt_slice = slice_pair["gt"]

omarsar / mt_libraries.sh

Created October 15, 2018 05:16

	!pip3 install http://download.pytorch.org/whl/cu80/torch-0.4.0-cp36-cp36m-linux_x86_64.whl
	!pip3 install torchvision
	!pip install medicaltorch
	!pip3 install numpy==1.14.1

omarsar / import_mtorch.py

Created October 14, 2018 00:53

	from collections import defaultdict
	import time
	import os

	import numpy as np

	from tqdm import tqdm

	from tensorboardX import SummaryWriter

omarsar / medical_torch.sh

Created October 14, 2018 00:31

pip install medicaltorch