dataeval.utils.data.batch_voc¶

dataeval.utils.data.batch_voc(dataset, model, batch_size=64, flatten_labels=False)¶

Iterates through the dataset to generate model embeddings and store labels

Note

Due to a bug with the VOCDetection dataset and DataLoaders, the batching is done manually

Parameters:¶

dataset : dataeval.utils.data.datasets.VOCDetection¶

model : torch.nn.Module¶

batch_size : int¶

flatten_labels : bool¶

Return type:¶

tuple[torch.Tensor, list[str] | list[list[str]]]