Lightly

Lightly is a data-curation and active-learning platform for computer vision. The LightlyOne platform exposes a REST API at https://api.lightly.ai for managing datasets, samples, embeddings, cloud datasources, selection / active-learning runs (the LightlyOne Worker), tags, and jobs. Lightly also maintains the open-source lightly self-supervised learning SDK, plus LightlyTrain and LightlyEdge products.

6 APIs 0 Features

AIComputer VisionData CurationActive LearningEmbeddings

APIs

Lightly Datasets API

Create, list, update, and delete LightlyOne datasets (image, video, crop, and embedding dataset types), look datasets up by name, query child datasets, and register dataset uplo...

Lightly Samples and Embeddings API

Manage samples within a dataset (create, list, read, update) with signed read / write URLs, and manage embeddings - list, delete, request signed CSV read / write URLs, mark embe...

Lightly Datasources API

Configure and inspect the cloud datasource backing a dataset (S3, GCS, Azure, or local), list raw data, metadata, and prediction files in the datasource, and track the processed...

Lightly Selection and Active Learning API

Register LightlyOne Workers and schedule compute-worker runs that execute a selection / active-learning configuration over a dataset, producing curated subsets via embeddings, m...

Lightly Jobs and Tags API

Poll asynchronous platform jobs by id, and manage dataset tags - create initial and named tags, delete tags, and export tagged subsets to ZIP, filename lists, Label Studio, Labe...

Lightly Self-Supervised Learning SDK (OSS)

The open-source (Apache-2.0) lightly Python library for self-supervised representation learning on images (SimCLR, MoCo, DINO, BYOL, and more). It is a pip-installable PyTorch S...

Collections

LightlyOne Platform API

OPEN

Pricing Plans

Lightly Ai Plans Pricing

4 plans

PLANS

Rate Limits

Lightly Ai Rate Limits

5 limits

RATE LIMITS

FinOps

Lightly Ai Finops

FINOPS

Resources

👥

GitHubOrganization

Sources

opencollection: 1.0.0
info:
  name: LightlyOne Platform API
  version: '1.0'
request:
  auth:
    type: apikey
    key: token
    value: '{{lightlyToken}}'
    in: query
items:
- info:
    name: Datasets
    type: folder
  items:
  - info:
      name: List all datasets
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets
    docs: List all datasets the user has access to.
  - info:
      name: Create a dataset
      type: http
    http:
      method: POST
      url: https://api.lightly.ai/v1/datasets
      body:
        type: json
        data: "{\n  \"name\": \"my-dataset\",\n  \"type\": \"Images\"\n}"
    docs: Create a new dataset.
  - info:
      name: Get a dataset by id
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets/{datasetId}
    docs: Get a dataset by id.
  - info:
      name: Delete a dataset by id
      type: http
    http:
      method: DELETE
      url: https://api.lightly.ai/v1/datasets/{datasetId}
    docs: Delete a dataset by id.
- info:
    name: Samples and Embeddings
    type: folder
  items:
  - info:
      name: List samples in a dataset
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets/{datasetId}/samples
    docs: List samples in a dataset.
  - info:
      name: Get sample image read URL
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets/{datasetId}/samples/{sampleId}/readurl?type=full
    docs: Get a signed read URL for a sample image.
  - info:
      name: List embeddings for a dataset
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets/{datasetId}/embeddings
    docs: List embeddings for a dataset.
  - info:
      name: Get embeddings CSV write URL
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets/{datasetId}/embeddings/writeCSVUrl?name=my-embeddings
    docs: Get a signed URL to upload an embeddings CSV.
  - info:
      name: Trigger 2D embeddings job
      type: http
    http:
      method: POST
      url: https://api.lightly.ai/v1/datasets/{datasetId}/embeddings/{embeddingId}/trigger2dEmbeddingsJob
      body:
        type: json
        data: "{\n  \"dimensionalityReductionMethod\": \"PCA\"\n}"
    docs: Trigger computation of a 2D projection of the embeddings.
- info:
    name: Datasources
    type: folder
  items:
  - info:
      name: Get datasource for a dataset
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets/{datasetId}/datasource
    docs: Get the datasource configured for a dataset.
  - info:
      name: List raw samples in datasource
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets/{datasetId}/datasource/list
    docs: List raw sample files available in the datasource.
- info:
    name: Selection and Active Learning
    type: folder
  items:
  - info:
      name: List registered workers
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/docker/worker/default
    docs: List registered LightlyOne Workers.
  - info:
      name: Schedule a worker run on a dataset
      type: http
    http:
      method: POST
      url: https://api.lightly.ai/v1/datasets/{datasetId}/docker/worker/schedule
      body:
        type: json
        data: "{\n  \"config\": {},\n  \"selectionConfig\": {\n    \"nSamples\": 1000,\n    \"strategies\": [\n      { \"\
          input\": { \"type\": \"EMBEDDINGS\" }, \"strategy\": { \"type\": \"DIVERSITY\" } }\n    ]\n  },\n  \"priority\"\
          : \"MID\"\n}"
    docs: Schedule a LightlyOne Worker selection / active-learning run on a dataset.
  - info:
      name: List worker runs
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/docker/runs
    docs: List LightlyOne Worker runs.
- info:
    name: Jobs and Tags
    type: folder
  items:
  - info:
      name: Get job status by id
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/jobs/{jobId}
    docs: Get the status of an asynchronous job by id.
  - info:
      name: Create a tag in a dataset
      type: http
    http:
      method: POST
      url: https://api.lightly.ai/v1/datasets/{datasetId}/tags
      body:
        type: json
        data: "{\n  \"name\": \"selected-subset\",\n  \"prevTagId\": \"{tagId}\",\n  \"bitMaskData\": \"0x0\"\n}"
    docs: Create a tag in a dataset.
  - info:
      name: Export tag to Label Studio tasks
      type: http
    http:
      method: GET
      url: https://api.lightly.ai/v1/datasets/{datasetId}/tags/{tagId}/export/LabelStudio/tasks
    docs: Export a tagged subset as Label Studio tasks.