Dimensionality Reduction

FiftyOne grid plugin to perform dimensionality reduction on a subset and compute visualization

Installation

fiftyone plugins download https://github.com/prernadh/dimensionality_reduction

Refer to the main README for more information about managing downloaded plugins and developing plugins locally.

Run Example

After installing this plugin, you can try the example yourself on the quickstart dataset.

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")
model = foz.load_zoo_model("clip-vit-base32-torch")
view = dataset.limit(4)
dataset.save_view("4_images", view)

embeddings = view.compute_embeddings(
    model=model,
    embeddings_field="clip_embeddings"
)

session = fo.launch_app(dataset)

Select the 4_images saved view from the View dropdown on the left of the App. Click on the Insights icon in the sample grid -> pass in the embeddings field name (clip_embeddings) and click execute. (We recommend that you schedule delegation as dimensionality reduction with UMAP scales with the number of samples in your view.) Finally you can open the embeddings panel and see the generated embedding visualization.