tlc.client.torch.metrics.metrics_collectors.segmentation_metrics_collector

Collect metrics for segmentation problems.

Module Contents

Classes

Class	Description
SegmentationMetricsCollector	Collect predicted masks from model output.

Data

Data	Description
PREDICTED_MASK_METRIC_NAME	Key for a column containing predicted masks (Deprecated: use PREDICTED_MASK instead)

API

tlc.client.torch.metrics.metrics_collectors.segmentation_metrics_collector.PREDICTED_MASK_METRIC_NAME = None

Key for a column containing predicted masks (Deprecated: use PREDICTED_MASK instead)

class tlc.client.torch.metrics.metrics_collectors.segmentation_metrics_collector.SegmentationMetricsCollector(label_map: dict[int, str], preprocess_fn: Callable[[tlc.core.builtins.types.SampleData, tlc.client.torch.metrics.predictor.PredictorOutput], tuple[Any, Any]] | None = None)

Bases: tlc.client.torch.metrics.metrics_collectors.metrics_collector_base.MetricsCollector

Collect predicted masks from model output.

Predicted masks are converted to PIL images, which can be written to the Run folder by a MetricsTableWriter.

Initialize the SegmentationMetricsCollector.

Parameters:

• label_map – A dictionary mapping class ids to class labels.

• preprocess_fn – A function that pre-processes the model output before computing metrics.

compute_metrics(batch: tlc.core.builtins.types.SampleData, predictor_output: tlc.client.torch.metrics.predictor.PredictorOutput) → dict[str, tlc.core.builtins.types.MetricData]

Convert predicted masks from model output to PIL images.

The result of preprocessing the model output is expected to be a list of tensors.

Parameters:

• batch – The input batch (not used).

• predictor_output – The output from the Predictor.

Returns:

A batch of metrics, where each metric is a PIL image corresponding to a mask.

preprocess(batch: tlc.core.builtins.types.SampleData, predictor_output: tlc.client.torch.metrics.predictor.PredictorOutput) → tuple[tlc.core.builtins.types.SampleData, torch.Tensor]

Default preprocessor for segmentation output.

By default just forwards the model predictions.

Parameters:

• batch – A batch of samples.

• predictor_output – A batch of predictions.

Returns:

A tuple containing the preprocessed batch and predictions.

tensor_to_pil_image(predicted_mask: torch.Tensor) → PIL.Image.Image

property column_schemas: dict[str, tlc.core.schema.Schema]