Drawing Labels on Samples

FiftyOne provides native support for rendering annotated versions of image and video samples with label fields overlaid on the source media.

Basic recipe

The interface for drawing labels on samples is exposed via the Python library and the CLI. You can easily annotate one or more label fields on entire datasets or arbitrary subsets of your datasets that you have identified by constructing a DatasetView.

PythonCLI

You can draw labels on a collection of samples via the Dataset.draw_labels() and DatasetView.draw_labels() methods:

import fiftyone as fo

# The Dataset or DatasetView containing the samples you wish to draw
dataset_or_view = fo.load_dataset(...)

# The directory to which to write the annotated media
output_dir = "/path/for/output"

# The list of `Label` fields containing the labels that you wish to render on
# the source media (e.g., classifications or detections)
label_fields = ["ground_truth", "predictions"]  # for example

# Render the labels!
dataset_or_view.draw_labels(output_dir, label_fields=label_fields)

Examples

Drawing labels on images

The following snippet renders the ground truth and predicted labels on a few samples from the quickstart dataset:

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart", max_samples=10)

anno_image_paths = dataset.draw_labels(
    "/tmp/quickstart/draw-labels",
    label_fields=None,                  # all label fields
    # label_fields=["predictions"],     # only predictions
)
print(anno_image_paths)

Drawing labels on videos

The following snippet renders both sample-level and frame-level labels on a few videos from the quickstart-video dataset:

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart-video", max_samples=2).clone()

# Add some temporal detections
sample1 = dataset.first()
sample1["events"] = fo.TemporalDetections(
    detections=[
        fo.TemporalDetection(label="first", support=[31, 60]),
        fo.TemporalDetection(label="second", support=[90, 120]),
    ]
)
sample1.save()

sample2 = dataset.last()
sample2["events"] = fo.TemporalDetections(
    detections=[
        fo.TemporalDetection(label="first", support=[16, 45]),
        fo.TemporalDetection(label="second", support=[75, 104]),
    ]
)
sample2.save()

anno_video_paths = dataset.draw_labels(
    "/tmp/quickstart-video/draw-labels",
    label_fields=None,                      # all sample and frame labels
    # label_fields=["events"],              # only sample-level labels
    # label_fields=["frames.detections"],   # only frame-level labels
)
print(anno_video_paths)

Individual samples

You can also render annotated versions of individual samples directly by using the various methods exposed in the fiftyone.utils.annotations module.

For example, you can render an annotated version of an image sample with Classification and Detections labels overlaid via draw_labeled_image():

import fiftyone as fo
import fiftyone.utils.annotations as foua

# Example data
sample = fo.Sample(
    filepath="~/fiftyone/coco-2017/validation/data/000000000632.jpg",
    gt_label=fo.Classification(label="bedroom"),
    pred_label=fo.Classification(label="house", confidence=0.95),
    gt_objects=fo.Detections(
        detections=[
            fo.Detection(
                label="bed",
                bounding_box=[0.00510938, 0.55248447, 0.62692188, 0.43115942],
            ),
            fo.Detection(
                label="chair",
                bounding_box=[0.38253125, 0.47712215, 0.16362500, 0.18155280],
            ),
        ]
    ),
    pred_objects=fo.Detections(
        detections=[
            fo.Detection(
                label="bed",
                bounding_box=[0.10, 0.63, 0.50, 0.35],
                confidence=0.74,
            ),
            fo.Detection(
                label="chair",
                bounding_box=[0.39, 0.53, 0.15, 0.13],
                confidence=0.92,
            ),
        ]
    ),
)

# The path to write the annotated image
outpath = "/path/for/image-annotated.jpg"

# Render the annotated image
foua.draw_labeled_image(sample, outpath)

Similarly, you can draw an annotated version of a video sample with its frame labels overlaid via draw_labeled_video().

Customizing label rendering

You can customize the look-and-feel of the labels rendered by FiftyOne by providing a custom DrawConfig to the relevant drawing method, such as SampleCollection.draw_labels() or the underlying methods in the fiftyone.utils.annotations module.

Consult the DrawConfig docs for a complete description of the available parameters.

For example, the snippet below increases the font size and line thickness of the labels in the example above and includes the confidence of the predictions:

# Continuing from example above...

# Customize annotation rendering
config = foua.DrawConfig(
    {
        "font_size": 24,
        "bbox_linewidth": 5,
        "show_all_confidences": True,
        "per_object_label_colors": False,
    }
)

# Render the annotated image
foua.draw_labeled_image(sample, outpath, config=config)