Level 12: Framework Conversion

Datumaro allows seamless conversion of datasets to popular deep learning frameworks, such as PyTorch and TensorFlow. This is particularly useful when you are working with a dataset that needs to be used across different frameworks without manual reformatting.

Datumaro provides the FrameworkConverter class, which can be used to convert a dataset for various tasks like classification, detection, and segmentation.

Supported Tasks

• Classification

• Multilabel Classification

• Detection

• Instance Segmentation

• Semantic Segmentation

• Tabular Data

Python

With the PyTorch framework, you can convert a Datumaro dataset like this:

from datumaro.plugins.framework_converter import FrameworkConverter
from torchvision import transforms

transform = transforms.Compose([transforms.ToTensor()])
dm_dataset = ...  # Load your dataset here

First, we have to specify the dataset, subset, and task

multi_framework_dataset = FrameworkConverter(dm_dataset, subset="train", task="classification")
train_dataset = multi_framework_dataset.to_framework(framework="torch", transform=transform)

Through this, we convert the dataset to PyTorch format

from torch.utils.data import DataLoader
train_loader = DataLoader(train_dataset, batch_size=32)

Now we can use the train_dataset with PyTorch DataLoader

In this example:

• subset=”train” indicates that we are working with the training portion of the dataset.

• task=”classification” specifies that this is a classification task.

• The dataset is converted to PyTorch-compatible format using the to_framework method.