Run in Google Colab

View source on GitHub

Download notebook

# Drawing Labels on Samples This recipe demonstrates how to use FiftyOne to render annotated versions of image and video [samples](https://voxel51.com/docs/fiftyone/user_guide/using_datasets.html#samples) with their [label field(s)](https://voxel51.com/docs/fiftyone/user_guide/using_datasets.html#labels) overlaid. ## Setup If you haven’t already, install FiftyOne: In this recipe we’ll use the [FiftyOne Dataset Zoo](https://docs.voxel51.com/dataset_zoo/index.html) to download some labeled datasets to use as sample data for drawing labels. Behind the scenes, FiftyOne uses either the [TensorFlow Datasets](https://www.tensorflow.org/datasets) or [TorchVision Datasets](https://pytorch.org/vision/stable/datasets.html) libraries to wrangle the datasets, depending on which ML library you have installed. You can, for example, install PyTorch as follows (we’ll also need `pycocotools` to load the COCO dataset, in particular): ## Drawing COCO detections You can download the validation split of the COCO-2017 dataset to `~/fiftyone/coco-2017/validation` by running the following command: Now let’s load the dataset, extract a [DatasetView](https://voxel51.com/docs/fiftyone/user_guide/using_datasets.html#datasetviews) that contains 100 images from the dataset, and render them as annotated images with their ground truth labels overlaid: Let’s list the output directory to verify that the annotations have been generated: Here’s an example of an annotated image that was generated:  ## Drawing Caltech 101 classifications You can download the test split of the Caltech 101 dataset to `~/fiftyone/caltech101/test` by running the following command: Now let’s load the dataset, extract a [DatasetView](https://voxel51.com/docs/fiftyone/user_guide/using_datasets.html#datasetviews) that contains 100 images from the dataset, and render them as annotated images with their ground truth labels overlaid: Let’s list the output directory to verify that the annotations have been generated: Here’s an example of an annotated image that was generated:  ## Drawing labels on videos FiftyOne can also render frame labels onto video samples. To demonstrate, let’s work with the (small) video quickstart dataset from the zoo: Note that the dataset contains frame-level detections in the `objs` field of each frame. Let’s make a [DatasetView](https://voxel51.com/docs/fiftyone/user_guide/using_datasets.html#datasetviews) that contains a couple random videos from the dataset and render them as annotated videos with the frame-level detections overlaid: Let’s list the output directory to verify that the annotations have been generated: Here’s a snippet of an annotated video that was generated:  ## Cleanup You can cleanup the files generated by this recipe by running the command below: