Pinecone Integration

Pinecone is one of the most popular vector search engines available, and we’ve made it easy to use Pinecone’s vector search capabilities on your computer vision data directly from FiftyOne!

Follow these simple instructions to configure your credentials and get started using Pinecone + FiftyOne.

FiftyOne provides an API to create Pinecone indexes, upload vectors, and run similarity queries, both programmatically in Python and via point-and-click in the App.

Note

Did you know? You can search by natural language using Pinecone similarity indexes!

Basic recipe

The basic workflow to use Pinecone to create a similarity index on your FiftyOne datasets and use this to query your data is as follows:

1. Load a dataset into FiftyOne

2. Compute embedding vectors for samples or patches in your dataset, or select a model to use to generate embeddings

3. Use the compute_similarity() methodto generate a Pinecone similarity index for the samples or object patches in a dataset by setting the parameter backend="pinecone" and specifying a brain_key of your choice

4. Use this Pinecone similarity index to query your data with sort_by_similarity()

5. If desired, delete the index

The example below demonstrates this workflow.

Note

You must create a Pinecone account, download a Pinecone API key, and install the Pinecone Python client to run this example:

pip install -U pinecone-client

Note that you can store your Pinecone credentials as described in this section to avoid entering them manually each time you interact with your Pinecone index.

First let’s load a dataset into FiftyOne and compute embeddings for the samples:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

# Step 1: Load your data into FiftyOne
dataset = foz.load_zoo_dataset("quickstart")

# Steps 2 and 3: Compute embeddings and create a similarity index
pinecone_index = fob.compute_similarity(
    dataset,
    brain_key="pinecone_index",
    backend="pinecone",
)

Once the similarity index has been generated, we can query our data in FiftyOne by specifying the brain_key:

# Step 4: Query your data
query = dataset.first().id  # query by sample ID
view = dataset.sort_by_similarity(
    query,
    brain_key=brain_key,
    k=10,  # limit to 10 most similar samples
)

# Step 5 (optional): Cleanup

# Delete the Pinecone index
pinecone_index = dataset.load_brain_results(brain_key)
pinecone_index.cleanup()

# Delete run record from FiftyOne
dataset.delete_brain_run("pinecone_index")

Note

Skip to this section to see a variety of common Pinecone query patterns.

Setup

The easiest way to get started with Pinecone is to create a free Pinecone account and copy your Pinecone API key.

Installing the Pinecone client

In order to use the Pinecone backend, you must install the Pinecone Python client:

pip install pinecone-client

Using the Pinecone backend

By default, calling compute_similarity() or sort_by_similarity() will use an sklearn backend.

To use the Pinecone backend, simply set the optional backend parameter of compute_similarity() to "pinecone":

import fiftyone.brain as fob

fob.compute_similarity(..., backend="pinecone", ...)

Alternatively, you can permanently configure FiftyOne to use the Pinecone backend by setting the following environment variable:

export FIFTYONE_BRAIN_DEFAULT_SIMILARITY_BACKEND=pinecone

or by setting the default_similarity_backend parameter of your brain config located at ~/.fiftyone/brain_config.json:

{
    "default_similarity_backend": "pinecone"
}

Authentication

In order to connect to a Pinecone server, you must provide your credentials, which can be done in a variety of ways.

Environment variables (recommended)

The recommended way to configure your Pinecone credentials is to store them in the environment variables shown below, which are automatically accessed by FiftyOne whenever a connection to Pinecone is made:

export FIFTYONE_BRAIN_SIMILARITY_PINECONE_API_KEY=XXXXXX

# Serverless indexes
export FIFTYONE_BRAIN_SIMILARITY_PINECONE_CLOUD="aws"
export FIFTYONE_BRAIN_SIMILARITY_PINECONE_REGION="us-east-1"

# Pod-based indexes
export FIFTYONE_BRAIN_SIMILARITY_PINECONE_ENVIRONMENT="us-east-1-aws"

FiftyOne Brain config

You can also store your credentials in your brain config located at ~/.fiftyone/brain_config.json:

{
    "similarity_backends": {
        "pinecone": {
            "api_key": "XXXXXXXXXXXX",
            "cloud": "aws",                 # serverless indexes
            "region": "us-east-1",          # serverless indexes
            "environment": "us-east-1-aws"  # pod-based indexes
        }
    }
}

Note that this file will not exist until you create it.

Keyword arguments

You can manually provide your Pinecone credentials as keyword arguments each time you call methods like compute_similarity() that require connections to Pinecone:

import fiftyone.brain as fob

pinecone_index = fob.compute_similarity(
    ...
    backend="pinecone",
    brain_key="pinecone_index",
    api_key="XXXXXX",
    cloud="aws",
    region="us-east-1",
)

Note that, when using this strategy, you must manually provide the credentials when loading an index later via load_brain_results():

pinecone_index = dataset.load_brain_results(
    "pinecone_index",
    api_key="XXXXXX",
    cloud="aws",
    region="us-east-1",
)

Pinecone config parameters

The Pinecone backend supports a variety of query parameters that can be used to customize your similarity queries. These parameters include:

• index_name (None): the name of the Pinecone index to use or create. If not specified, a new unique name is generated automatically

• index_type (None): the index type to use when creating a new index. The supported values are ["serverless", "pod"], and the default is "serverless"

• namespace (None): a namespace under which to store vectors added to the index

• metric (“cosine”): the distance/similarity metric to use for the index. Supported values are ("cosine", "dotproduct", "euclidean")

• replicas (None): an optional number of replicas to use when creating a new pod-based index

• shards (None): an optional number of shards to use when creating a new pod-based index

• pods (None): an optional number of pods to use when creating a new pod-based index

• pod_type (None): an optional pod type to use when creating a new pod-based index

For detailed information on these parameters, see the Pinecone documentation.

You can specify these parameters via any of the strategies described in the previous section. Here’s an example of a brain config that configures a serverless index:

{
    "similarity_backends": {
        "pinecone": {
            "index_name": "your-index",
            "index_type": "serverless",
            "metric": "cosine",
        }
    }
}

However, typically these parameters are directly passed to compute_similarity() to configure a specific new index:

pinecone_index = fob.compute_similarity(
    ...
    backend="pinecone",
    brain_key="pinecone_index",
    index_name="your-index",
    index_type="serverless",
    metric="cosine",
)

Managing brain runs

FiftyOne provides a variety of methods that you can use to manage brain runs.

For example, you can call list_brain_runs() to see the available brain keys on a dataset:

import fiftyone.brain as fob

# List all brain runs
dataset.list_brain_runs()

# Only list similarity runs
dataset.list_brain_runs(type=fob.Similarity)

# Only list specific similarity runs
dataset.list_brain_runs(
    type=fob.Similarity,
    patches_field="ground_truth",
    supports_prompts=True,
)

Or, you can use get_brain_info() to retrieve information about the configuration of a brain run:

info = dataset.get_brain_info(brain_key)
print(info)

Use load_brain_results() to load the SimilarityIndex instance for a brain run.

You can use rename_brain_run() to rename the brain key associated with an existing similarity results run:

dataset.rename_brain_run(brain_key, new_brain_key)

Finally, you can use delete_brain_run() to delete the record of a similarity index computation from your FiftyOne dataset:

dataset.delete_brain_run(brain_key)

Note

Calling delete_brain_run() only deletes the record of the brain run from your FiftyOne dataset; it will not delete any associated Pinecone index, which you can do as follows:

# Delete the Pinecone index
pinecone_index = dataset.load_brain_results(brain_key)
pinecone_index.cleanup()

Examples

This section demonstrates how to perform some common vector search workflows on a FiftyOne dataset using the Pinecone backend.

Note

All of the examples below assume you have configured your Pinecone API key as described in this section.

Create a similarity index

In order to create a new Pinecone similarity index, you need to specify either the embeddings or model argument to compute_similarity(). Here’s a few possibilities:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")
model_name = "clip-vit-base32-torch"
model = foz.load_zoo_model(model_name)
brain_key = "pinecone_index"

# Option 1: Compute embeddings on the fly from model name
fob.compute_similarity(
    dataset,
    model=model_name,
    backend="pinecone",
    brain_key=brain_key,
)

# Option 2: Compute embeddings on the fly from model instance
fob.compute_similarity(
    dataset,
    model=model,
    backend="pinecone",
    brain_key=brain_key,
)

# Option 3: Pass precomputed embeddings as a numpy array
embeddings = dataset.compute_embeddings(model)
fob.compute_similarity(
    dataset,
    embeddings=embeddings,
    backend="pinecone",
    brain_key=brain_key,
)

# Option 4: Pass precomputed embeddings by field name
dataset.compute_embeddings(model, embeddings_field="embeddings")
fob.compute_similarity(
    dataset,
    embeddings="embeddings",
    backend="pinecone",
    brain_key=brain_key,
)

Note

You can customize the Pinecone index by passing any supported parameters as extra kwargs.

Create a patch similarity index

You can also create a similarity index for object patches within your dataset by specifying a patches_field argument to compute_similarity():

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

fob.compute_similarity(
    dataset,
    patches_field="ground_truth",
    model="clip-vit-base32-torch",
    backend="pinecone",
    brain_key="pinecone_patches",
)

Note

You can customize the Pinecone index by passing any supported parameters as extra kwargs.

Connect to an existing index

If you have already created a Pinecone index storing the embedding vectors for the samples or patches in your dataset, you can connect to it by passing the index_name to compute_similarity():

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",      # zoo model used (if applicable)
    embeddings=False,                   # don't compute embeddings
    index_name="your-index",            # the existing Pinecone index
    brain_key="pinecone_index",
    backend="pinecone",
)

Add/remove embeddings from an index

You can use add_to_index() and remove_from_index() to add and remove embeddings from an existing Pinecone index.

These methods can come in handy if you modify your FiftyOne dataset and need to update the Pinecone index to reflect these changes:

import numpy as np

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

pinecone_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    brain_key="pinecone_index",
    backend="pinecone",
)
print(pinecone_index.total_index_size)  # 200

view = dataset.take(10)
ids = view.values("id")

# Delete 10 samples from a dataset
dataset.delete_samples(view)

# Delete the corresponding vectors from the index
pinecone_index.remove_from_index(sample_ids=ids)

# Add 20 samples to a dataset
samples = [fo.Sample(filepath="tmp%d.jpg" % i) for i in range(20)]
sample_ids = dataset.add_samples(samples)

# Add corresponding embeddings to the index
embeddings = np.random.rand(20, 512)
pinecone_index.add_to_index(embeddings, sample_ids)

print(pinecone_index.total_index_size)  # 210

Retrieve embeddings from an index

You can use get_embeddings() to retrieve embeddings from a Pinecone index by ID:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

pinecone_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    brain_key="pinecone_index",
    backend="pinecone",
)

# Retrieve embeddings for the entire dataset
ids = dataset.values("id")
embeddings, sample_ids, _ = pinecone_index.get_embeddings(sample_ids=ids)
print(embeddings.shape)  # (200, 512)
print(sample_ids.shape)  # (200,)

# Retrieve embeddings for a view
ids = dataset.take(10).values("id")
embeddings, sample_ids, _ = pinecone_index.get_embeddings(sample_ids=ids)
print(embeddings.shape)  # (10, 512)
print(sample_ids.shape)  # (10,)

Querying a Pinecone index

You can query a Pinecone index by appending a sort_by_similarity() stage to any dataset or view. The query can be any of the following:

• An ID (sample or patch)

• A query vector of same dimension as the index

• A list of IDs (samples or patches)

• A text prompt (if supported by the model)

import numpy as np

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    brain_key="pinecone_index",
    backend="pinecone",
)

# Query by vector
query = np.random.rand(512)  # matches the dimension of CLIP embeddings
view = dataset.sort_by_similarity(query, k=10, brain_key="pinecone_index")

# Query by sample ID
query = dataset.first().id
view = dataset.sort_by_similarity(query, k=10, brain_key="pinecone_index")

# Query by a list of IDs
query = [dataset.first().id, dataset.last().id]
view = dataset.sort_by_similarity(query, k=10, brain_key="pinecone_index")

# Query by text prompt
query = "a photo of a dog"
view = dataset.sort_by_similarity(query, k=10, brain_key="pinecone_index")

Note

Performing a similarity search on a DatasetView will only return results from the view; if the view contains samples that were not included in the index, they will never be included in the result.

This means that you can index an entire Dataset once and then perform searches on subsets of the dataset by constructing views that contain the images of interest.

Accessing the Pinecone client

You can use the index property of a Pinecone index to directly access the underlying Pinecone client instance and use its methods as desired:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

pinecone_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    brain_key="pinecone_index",
    backend="pinecone",
)

print(pinecone_index.index)

Advanced usage

As previously mentioned, you can customize your Pinecone indexes by providing optional parameters to compute_similarity().

Here’s an example of creating a similarity index backed by a customized Pinecone index. Just for fun, we’ll specify a custom index name, use dot product similarity, and populate the index for only a subset of our dataset:

import fiftyone as fo
import fiftyone.brain as fob
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")

# Create a custom Pinecone index
pinecone_index = fob.compute_similarity(
    dataset,
    model="clip-vit-base32-torch",
    embeddings=False,  # we'll add embeddings below
    metric="dotproduct",
    brain_key="pinecone_index",
    backend="pinecone",
    index_name="custom-pinecone-index",
)

# Add embeddings for a subset of the dataset
view = dataset.take(10)
embeddings, sample_ids, _ = pinecone_index.compute_embeddings(view)
pinecone_index.add_to_index(embeddings, sample_ids)

print(pinecone_index.index)