from __future__ import print_function import os,sys,inspect currentdir = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe()))) parentdir = os.path.dirname(os.path.dirname(currentdir)) sys.path.insert(0,parentdir) from utils.args import args import setup.categories.ae_setup as AESetup from models.autoencoders import * from datasets.NIH_Chest import NIHChestBinaryTrainSplit if __name__ == "__main__": dataset = NIHChestBinaryTrainSplit(root_path=os.path.join(args.root_path, "NIHCC"), binary=True, expand_channels=False, downsample=64) model = Generic_VAE(dims=(1, 64, 64), max_channels=512, depth=12, n_hidden=512) #model = ALILikeVAE(dims=(1, 64, 64)) AESetup.train_variational_autoencoder(args, model=model, dataset=dataset.get_D1_train(), BCE_Loss=False)
from __future__ import print_function import os, sys, inspect currentdir = os.path.dirname( os.path.abspath(inspect.getfile(inspect.currentframe()))) parentdir = os.path.dirname(os.path.dirname(currentdir)) sys.path.insert(0, parentdir) from utils.args import args import setup.categories.ae_setup as AESetup from models.autoencoders import Generic_VAE, Generic_AE, Residual_AE from datasets.NIH_Chest import NIHChestBinaryTrainSplit if __name__ == "__main__": dataset = NIHChestBinaryTrainSplit(root_path=os.path.join( args.root_path, "NIHCC"), binary=True, expand_channels=False, downsample=64) model = Residual_AE(dims=(1, 64, 64)) AESetup.train_autoencoder(args, model=model, dataset=dataset.get_D1_train(), BCE_Loss=False)
currentdir = os.path.dirname( os.path.abspath(inspect.getfile(inspect.currentframe()))) parentdir = os.path.dirname(os.path.dirname(currentdir)) sys.path.insert(0, parentdir) import models as Models import global_vars as Global from utils.args import args import categories.classifier_setup as CLSetup from models.classifiers import NIHDenseBinary, NIHChestVGG from datasets.NIH_Chest import NIHChestBinaryTrainSplit if __name__ == "__main__": dataset = NIHChestBinaryTrainSplit(root_path=os.path.join( args.root_path, "NIHCC"), binary=True) model = NIHChestVGG() CLSetup.train_classifier(args, model=model, dataset=dataset.get_D1_train()) # task_list = [ # # The list of models, The function that does the training, Can I skip-test?, suffix of the operation. # # The procedures that can be skip-test are the ones that we can determine # # whether we have done them before without instantiating the network architecture or dataset. # # saves quite a lot of time when possible. # (Global.dataset_reference_classifiers, CLSetup.train_classifier, True, ['base0']), # (Global.dataset_reference_classifiers, KLogisticSetup.train_classifier, True, ['KLogistic']), # (Global.dataset_reference_classifiers, DeepEnsembleSetup.train_classifier, True, ['DE.%d'%i for i in range(5)]), # (Global.dataset_reference_autoencoders, AESetup.train_BCE_AE, False, []), # (Global.dataset_reference_autoencoders, AESetup.train_MSE_AE, False, []), # (Global.dataset_reference_vaes, AESetup.train_variational_autoencoder, False, []),
'bceaeknn/1', 'vaemseaeknn/2', 'vaebceaeknn/2', 'mseaeknn/2', 'bceaeknn/2', 'vaemseaeknn/4', 'vaebceaeknn/4', 'mseaeknn/4', 'bceaeknn/4', 'vaemseaeknn/8', 'vaebceaeknn/8', 'mseaeknn/8', 'bceaeknn/8', ] D1 = NIHChestBinaryTrainSplit( root_path=os.path.join(args.root_path, 'NIHCC')) D164 = NIHChestBinaryTrainSplit(root_path=os.path.join( args.root_path, 'NIHCC'), downsample=64) args.D1 = 'NIHCC' #Usecase 1 Evaluation d2s = [ 'CIFAR10', 'UniformNoise', 'MURAHAND', ] D2s = [] for d2 in d2s: dataset = Global.all_datasets[d2] if 'dataset_path' in dataset.__dict__:
from __future__ import print_function import os, sys, inspect from termcolor import colored import torch currentdir = os.path.dirname( os.path.abspath(inspect.getfile(inspect.currentframe()))) parentdir = os.path.dirname(os.path.dirname(currentdir)) sys.path.insert(0, parentdir) import models as Models import global_vars as Global from utils.args import args import categories.classifier_setup as CLSetup from models.classifiers import NIHDenseBinary from datasets.NIH_Chest import NIHChestBinaryTrainSplit if __name__ == "__main__": dataset = NIHChestBinaryTrainSplit(root_path=os.path.join( args.root_path, "NIHCC"), binary=True) model = NIHDenseBinary("mono_model.pth.tar") CLSetup.train_classifier(args, model=model, dataset=dataset.get_D1_train(), balanced=True)