Deep Learning with PyTorch

Creating a custom dataset along with data loaders

We will not be able to use the FeaturesDataset class as it is, since it was developed to pick from the output of only one model. So, the following implementation contains minor changes to the FeaturesDataset class to accommodate all the three different generated features:

class FeaturesDataset(Dataset):        def __init__(self,featlst1,featlst2,featlst3,labellst):        self.featlst1 = featlst1        self.featlst2 = featlst2        self.featlst3 = featlst3        self.labellst = labellst            def __getitem__(self,index):        return (self.featlst1[index],self.featlst2[index],self.featlst3[index],self.labellst[index])        def __len__(self):        return len(self.labellst)trn_feat_dset = FeaturesDataset(trn_resnet_features,trn_inception_features.features,trn_densenet_features,trn_labels) ...